CINXE.COM

<!DOCTYPE html> <html lang="en"> <head> <meta content="text/html; charset=utf-8" http-equiv="content-type"/> <title>When Pigs Get Sick: Multi-Agent AI for Swine Disease Detection</title>  <meta content="width=device-width, initial-scale=1, shrink-to-fit=no" name="viewport"/> <link href="https://cdn.jsdelivr.net/npm/bootstrap@5.3.0/dist/css/bootstrap.min.css" rel="stylesheet" type="text/css"/> <link href="/static/browse/0.3.4/css/ar5iv.0.7.9.min.css" rel="stylesheet" type="text/css"/> <link href="/static/browse/0.3.4/css/ar5iv-fonts.0.7.9.min.css" rel="stylesheet" type="text/css"/> <link href="/static/browse/0.3.4/css/latexml_styles.css" rel="stylesheet" type="text/css"/> <script src="https://cdn.jsdelivr.net/npm/bootstrap@5.3.0/dist/js/bootstrap.bundle.min.js"></script> <script src="https://cdnjs.cloudflare.com/ajax/libs/html2canvas/1.3.3/html2canvas.min.js"></script> <script src="/static/browse/0.3.4/js/addons_new.js"></script> <script src="/static/browse/0.3.4/js/feedbackOverlay.js"></script> <base href="/html/2503.15204v1/"/></head> <body> <nav class="ltx_page_navbar"> <nav class="ltx_TOC"> <ol class="ltx_toclist"> <li class="ltx_tocentry ltx_tocentry_section"><a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#S1" title="In When Pigs Get Sick: Multi-Agent AI for Swine Disease Detection"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">1 </span>Introduction</span></a></li> <li class="ltx_tocentry ltx_tocentry_section"><a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#S2" title="In When Pigs Get Sick: Multi-Agent AI for Swine Disease Detection"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">2 </span>Related Works</span></a></li> <li class="ltx_tocentry ltx_tocentry_section"> <a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#S3" title="In When Pigs Get Sick: Multi-Agent AI for Swine Disease Detection"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">3 </span>Methodology</span></a> <ol class="ltx_toclist ltx_toclist_section"> <li class="ltx_tocentry ltx_tocentry_subsection"><a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#S3.SS1" title="In 3 Methodology ‣ When Pigs Get Sick: Multi-Agent AI for Swine Disease Detection"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">3.1 </span>Query Classification</span></a></li> <li class="ltx_tocentry ltx_tocentry_subsection"> <a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#S3.SS2" title="In 3 Methodology ‣ When Pigs Get Sick: Multi-Agent AI for Swine Disease Detection"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">3.2 </span>Multi-Agent Disease Prediction</span></a> <ol class="ltx_toclist ltx_toclist_subsection"> <li class="ltx_tocentry ltx_tocentry_subsubsection"><a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#S3.SS2.SSS1" title="In 3.2 Multi-Agent Disease Prediction ‣ 3 Methodology ‣ When Pigs Get Sick: Multi-Agent AI for Swine Disease Detection"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">3.2.1 </span>Stage-Wise Symptom Collection</span></a></li> <li class="ltx_tocentry ltx_tocentry_subsubsection"><a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#S3.SS2.SSS2" title="In 3.2 Multi-Agent Disease Prediction ‣ 3 Methodology ‣ When Pigs Get Sick: Multi-Agent AI for Swine Disease Detection"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">3.2.2 </span>State Transitions</span></a></li> <li class="ltx_tocentry ltx_tocentry_subsubsection"><a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#S3.SS2.SSS3" title="In 3.2 Multi-Agent Disease Prediction ‣ 3 Methodology ‣ When Pigs Get Sick: Multi-Agent AI for Swine Disease Detection"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">3.2.3 </span>Confidence-Based Disease Prediction</span></a></li> </ol> </li> <li class="ltx_tocentry ltx_tocentry_subsection"> <a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#S3.SS3" title="In 3 Methodology ‣ When Pigs Get Sick: Multi-Agent AI for Swine Disease Detection"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">3.3 </span>Recommendations</span></a> <ol class="ltx_toclist ltx_toclist_subsection"> <li class="ltx_tocentry ltx_tocentry_subsubsection"><a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#S3.SS3.SSS1" title="In 3.3 Recommendations ‣ 3 Methodology ‣ When Pigs Get Sick: Multi-Agent AI for Swine Disease Detection"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">3.3.1 </span>Algorithmic Framework</span></a></li> <li class="ltx_tocentry ltx_tocentry_subsubsection"><a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#S3.SS3.SSS2" title="In 3.3 Recommendations ‣ 3 Methodology ‣ When Pigs Get Sick: Multi-Agent AI for Swine Disease Detection"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">3.3.2 </span>Integrated Recommendation Pipeline</span></a></li> </ol> </li> </ol> </li> <li class="ltx_tocentry ltx_tocentry_section"> <a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#S4" title="In When Pigs Get Sick: Multi-Agent AI for Swine Disease Detection"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">4 </span>Experimental Setup</span></a> <ol class="ltx_toclist ltx_toclist_section"> <li class="ltx_tocentry ltx_tocentry_subsection"><a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#S4.SS1" title="In 4 Experimental Setup ‣ When Pigs Get Sick: Multi-Agent AI for Swine Disease Detection"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">4.1 </span>Data Collection</span></a></li> <li class="ltx_tocentry ltx_tocentry_subsection"> <a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#S4.SS2" title="In 4 Experimental Setup ‣ When Pigs Get Sick: Multi-Agent AI for Swine Disease Detection"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">4.2 </span>Evaluation Methodology</span></a> <ol class="ltx_toclist ltx_toclist_subsection"> <li class="ltx_tocentry ltx_tocentry_subsubsection"><a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#S4.SS2.SSS1" title="In 4.2 Evaluation Methodology ‣ 4 Experimental Setup ‣ When Pigs Get Sick: Multi-Agent AI for Swine Disease Detection"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">4.2.1 </span>Evaluation Metrics for Query Classification and Routing</span></a></li> <li class="ltx_tocentry ltx_tocentry_subsubsection"><a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#S4.SS2.SSS2" title="In 4.2 Evaluation Methodology ‣ 4 Experimental Setup ‣ When Pigs Get Sick: Multi-Agent AI for Swine Disease Detection"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">4.2.2 </span>Evaluation Metrics for Symptom-Based Diagnostic Queries</span></a></li> <li class="ltx_tocentry ltx_tocentry_subsubsection"><a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#S4.SS2.SSS3" title="In 4.2 Evaluation Methodology ‣ 4 Experimental Setup ‣ When Pigs Get Sick: Multi-Agent AI for Swine Disease Detection"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">4.2.3 </span>Evaluation Metrics for Knowledge Retrieval Queries</span></a></li> </ol> </li> </ol> </li> <li class="ltx_tocentry ltx_tocentry_section"> <a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#S5" title="In When Pigs Get Sick: Multi-Agent AI for Swine Disease Detection"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">5 </span>Results</span></a> <ol class="ltx_toclist ltx_toclist_section"> <li class="ltx_tocentry ltx_tocentry_subsection"><a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#S5.SS1" title="In 5 Results ‣ When Pigs Get Sick: Multi-Agent AI for Swine Disease Detection"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">5.1 </span>Query Classification Results</span></a></li> <li class="ltx_tocentry ltx_tocentry_subsection"><a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#S5.SS2" title="In 5 Results ‣ When Pigs Get Sick: Multi-Agent AI for Swine Disease Detection"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">5.2 </span>Symptom-Based Diagnostic Results</span></a></li> <li class="ltx_tocentry ltx_tocentry_subsection"><a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#S5.SS3" title="In 5 Results ‣ When Pigs Get Sick: Multi-Agent AI for Swine Disease Detection"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">5.3 </span>Knowledge Retrieval Results</span></a></li> </ol> </li> <li class="ltx_tocentry ltx_tocentry_section"> <a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#S6" title="In When Pigs Get Sick: Multi-Agent AI for Swine Disease Detection"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">6 </span>Discussion</span></a> <ol class="ltx_toclist ltx_toclist_section"> <li class="ltx_tocentry ltx_tocentry_subsection"><a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#S6.SS1" title="In 6 Discussion ‣ When Pigs Get Sick: Multi-Agent AI for Swine Disease Detection"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">6.1 </span>Ambiguity in Query Classification</span></a></li> <li class="ltx_tocentry ltx_tocentry_subsection"><a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#S6.SS2" title="In 6 Discussion ‣ When Pigs Get Sick: Multi-Agent AI for Swine Disease Detection"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">6.2 </span>Misclassification of Diseases</span></a></li> <li class="ltx_tocentry ltx_tocentry_subsection"><a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#S6.SS3" title="In 6 Discussion ‣ When Pigs Get Sick: Multi-Agent AI for Swine Disease Detection"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">6.3 </span>Low Confidence Predictions</span></a></li> </ol> </li> <li class="ltx_tocentry ltx_tocentry_section"><a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#S7" title="In When Pigs Get Sick: Multi-Agent AI for Swine Disease Detection"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">7 </span>Conclusion and Future works</span></a></li> <li class="ltx_tocentry ltx_tocentry_section"><a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#S8" title="In When Pigs Get Sick: Multi-Agent AI for Swine Disease Detection"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">8 </span>Acknowledgments</span></a></li> </ol></nav> </nav> <div class="ltx_page_main"> <div class="ltx_page_content"> <article class="ltx_document ltx_authors_1line"> <h1 class="ltx_title ltx_title_document">When Pigs Get Sick: Multi-Agent AI <br class="ltx_break"/>for Swine Disease Detection</h1> <div class="ltx_authors"> <span class="ltx_creator ltx_role_author"> <span class="ltx_personname"> Tittaya Mairittha, Tanakon Sawanglok, Panuwit Raden, Sorrawit Treesuk <br class="ltx_break"/>AXONS <br class="ltx_break"/><span class="ltx_text ltx_font_typewriter" id="id1.1.id1">aimlstaff@axonstech.com</span> <br class="ltx_break"/> </span></span> </div> <div class="ltx_abstract"> <h6 class="ltx_title ltx_title_abstract">Abstract</h6> <p class="ltx_p" id="id2.id1">Swine disease surveillance is critical to the sustainability of global agriculture, yet its effectiveness is frequently undermined by limited veterinary resources, delayed identification of cases, and variability in diagnostic accuracy. To overcome these barriers, we introduce a novel AI-powered, multi-agent diagnostic system that leverages Retrieval-Augmented Generation (RAG) to deliver timely, evidence-based disease detection and clinical guidance. By automatically classifying user inputs into either Knowledge Retrieval Queries or Symptom-Based Diagnostic Queries, the system ensures targeted information retrieval and facilitates precise diagnostic reasoning. An adaptive questioning protocol systematically collects relevant clinical signs, while a confidence-weighted decision fusion mechanism integrates multiple diagnostic hypotheses to generate robust disease predictions and treatment recommendations. Comprehensive evaluations encompassing query classification, disease diagnosis, and knowledge retrieval demonstrate that the system achieves high accuracy, rapid response times, and consistent reliability. By providing a scalable, AI-driven diagnostic framework, this approach enhances veterinary decision-making, advances sustainable livestock management practices, and contributes substantively to the realization of global food security.</p> </div> <section class="ltx_section" id="S1"> <h2 class="ltx_title ltx_title_section"> <span class="ltx_tag ltx_tag_section">1 </span>Introduction</h2> <div class="ltx_para ltx_noindent" id="S1.p1"> <p class="ltx_p" id="S1.p1.1">Efficient surveillance and management of swine diseases are vital for maintaining global agricultural productivity, ensuring food security, and safeguarding animal welfare. However, the shortage of specialized veterinary expertise, especially in resource-limited regions, often results in delayed disease detection and ineffective responses. These challenges exacerbate economic losses, disrupt food supply chains, and negatively impact overall animal health <cite class="ltx_cite ltx_citemacro_cite">[<a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#bib.bib1" title="">1</a>]</cite>. Consequently, the timely and accurate identification of diseases is fundamental for the stability and sustainability of the swine industry.</p> </div> <div class="ltx_para ltx_noindent" id="S1.p2"> <p class="ltx_p" id="S1.p2.1">To address this critical expertise gap, we introduce an AI-powered multi-agent diagnostic platform that utilizes Retrieval-Augmented Generation (RAG). The system supports veterinarians, animal health professionals, and swine husbandry professionals by delivering precise and context-sensitive diagnostic insights. The diagnostic process begins with an advanced query-classification methodology, categorizing submitted queries into two primary types: Knowledge Retrieval Queries, addressing general swine health information such as preventive care, vaccination strategies, medication guidelines, and biosecurity best practices, and Symptom-Based Diagnostic Queries, requiring comprehensive analysis of symptom descriptions to generate accurate differential diagnoses and targeted treatment recommendations.</p> </div> <div class="ltx_para ltx_noindent" id="S1.p3"> <p class="ltx_p" id="S1.p3.1">A core innovation of our platform is its adaptive questioning mechanism, which dynamically collects detailed symptom data to refine diagnostic accuracy. Disease predictions are further enhanced through a confidence-weighted decision fusion approach, which prioritizes high-certainty results while flagging cases that require further expert evaluation. Additionally, the system provides actionable guidance, including disease control recommendations, sample collection protocols, treatment instructions, and preventive measures, ensuring its practical applicability in swine health management.</p> </div> <div class="ltx_para ltx_noindent" id="S1.p4"> <p class="ltx_p" id="S1.p4.1">The effectiveness of our diagnostic approach was validated through a rigorous evaluation framework integrating self-correcting Large Language Models (LLMs) to minimize hallucinations and utilizing a curated domain-specific corpus to improve diagnostic precision and contextual reliability. Performance benchmarking was conducted across four critical swine diseases: African Swine Fever (ASF), Porcine Reproductive and Respiratory Syndrome (PRRS), Porcine Epidemic Diarrhea (PED), and Foot-and-Mouth Disease (FMD). Continuous expert-driven feedback and iterative refinements significantly enhanced the system’s diagnostic accuracy, clinical relevance, and overall practical utility.</p> </div> <div class="ltx_para ltx_noindent" id="S1.p5"> <p class="ltx_p" id="S1.p5.1">By bridging the veterinary expertise gap with scalable AI-driven diagnostics, our platform empowers animal health professionals to make informed, timely decisions. This advancement strengthens disease surveillance, enhances animal welfare, and bolsters the resilience and sustainability of global pork production.</p> </div> </section> <section class="ltx_section" id="S2"> <h2 class="ltx_title ltx_title_section"> <span class="ltx_tag ltx_tag_section">2 </span>Related Works</h2> <div class="ltx_para ltx_noindent" id="S2.p1"> <p class="ltx_p" id="S2.p1.1">Recent advancements in LLMs have significantly contributed to the development of multi-agent AI architectures, particularly for complex decision-making in healthcare applications. For instance, MDAgents <cite class="ltx_cite ltx_citemacro_cite">[<a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#bib.bib2" title="">2</a>]</cite> dynamically adapt agent roles based on case complexity, while the AMSC framework <cite class="ltx_cite ltx_citemacro_cite">[<a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#bib.bib3" title="">3</a>]</cite> leverages probabilistic coordination among specialist agents to improve diagnostic accuracy. Additionally, researchers have explored deep reinforcement learning for personalized treatment optimization <cite class="ltx_cite ltx_citemacro_cite">[<a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#bib.bib4" title="">4</a>, <a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#bib.bib5" title="">5</a>]</cite>, ensemble learning for enhancing diagnostic precision <cite class="ltx_cite ltx_citemacro_cite">[<a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#bib.bib6" title="">6</a>]</cite>, and knowledge-based AI frameworks that integrate structured clinical guidelines into decision-support systems <cite class="ltx_cite ltx_citemacro_cite">[<a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#bib.bib7" title="">7</a>]</cite>. These developments highlight the superiority of multi-agent architectures over single-model AI systems, particularly in scalability, adaptability, and diagnostic consistency.</p> </div> <div class="ltx_para ltx_noindent" id="S2.p2"> <p class="ltx_p" id="S2.p2.1">In veterinary medicine, AI-powered diagnostic and decision-support systems are emerging, albeit at an earlier stage than in human healthcare. Sobkowich <cite class="ltx_cite ltx_citemacro_cite">[<a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#bib.bib8" title="">8</a>]</cite> emphasizes the increasing role of AI in veterinary clinical workflows, including automated disease detection and medical imaging analysis, demonstrating how AI-driven tools streamline and enhance veterinary practice. Beyond diagnostics, machine learning models assist veterinarians in personalizing treatment strategies based on an animal’s genetic profile, medical history, and clinical symptoms <cite class="ltx_cite ltx_citemacro_cite">[<a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#bib.bib9" title="">9</a>]</cite>. Meanwhile, Joslyn et al. <cite class="ltx_cite ltx_citemacro_cite">[<a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#bib.bib10" title="">10</a>]</cite> highlight the importance of collaborative AI models in veterinary radiology, arguing that multi-agent coordination—akin to human medical expert consultations—enhances diagnostic reliability in complex cases.</p> </div> <div class="ltx_para ltx_noindent" id="S2.p3"> <p class="ltx_p" id="S2.p3.1">AI innovations in veterinary applications extend to specialized LLMs trained on clinical notes, which have been shown to outperform traditional disease prediction models <cite class="ltx_cite ltx_citemacro_cite">[<a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#bib.bib11" title="">11</a>]</cite>. Similarly, generative AI models have been applied to prescription analysis, automating medication dosage extraction and reducing prescription errors <cite class="ltx_cite ltx_citemacro_cite">[<a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#bib.bib12" title="">12</a>]</cite>. Additionally, multimodal AI frameworks such as MM-Vet integrate textual, imaging, and sensor data, expanding beyond text-based diagnostics to comprehensive AI-assisted decision-making <cite class="ltx_cite ltx_citemacro_cite">[<a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#bib.bib13" title="">13</a>]</cite>. These advancements form a strong foundation for AI-driven decision support in veterinary medicine, with multi-agent methodologies poised to improve diagnostic accuracy, efficiency, and scalability.</p> </div> <div class="ltx_para ltx_noindent" id="S2.p4"> <p class="ltx_p" id="S2.p4.1">In livestock farming, AI-powered precision monitoring systems are transforming disease surveillance and herd management. Even in traditional farm environments, multi-agent AI frameworks combined with IoT-based sensor networks have been implemented for real-time remote health monitoring <cite class="ltx_cite ltx_citemacro_cite">[<a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#bib.bib14" title="">14</a>]</cite>. In swine disease management, AI-driven innovations are shaping next-generation multi-agent diagnostic systems. For example, deep learning models have been developed to interpret lateral flow assay results for ASF, providing rapid and sensitive field diagnostics for early outbreak detection <cite class="ltx_cite ltx_citemacro_cite">[<a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#bib.bib15" title="">15</a>]</cite>.</p> </div> <div class="ltx_para ltx_noindent" id="S2.p5"> <p class="ltx_p" id="S2.p5.1">AI-powered acoustic analysis has also been employed to automatically detect respiratory diseases in pigs based on cough patterns <cite class="ltx_cite ltx_citemacro_cite">[<a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#bib.bib16" title="">16</a>]</cite>. Notably, AI cough recognition models can differentiate between productive, disease-related coughs and non-pathological environmental coughs caused by dust or irritants <cite class="ltx_cite ltx_citemacro_cite">[<a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#bib.bib17" title="">17</a>]</cite>. Furthermore, AI-driven continuous cough monitoring, when combined with periodic pathogen testing, acts as an early-warning system for respiratory outbreaks. Field studies indicate that AI-monitored distress signals correlate with spikes in specific pathogens, enabling proactive disease containment before traditional detection methods confirm an outbreak <cite class="ltx_cite ltx_citemacro_cite">[<a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#bib.bib18" title="">18</a>]</cite>.</p> </div> <div class="ltx_para ltx_noindent" id="S2.p6"> <p class="ltx_p" id="S2.p6.1">While cough recognition AI focuses on respiratory disease detection, computer vision techniques provide an additional layer of disease surveillance by analyzing pig behavior and physiological changes. AI models analyzing video feeds have demonstrated the ability to detect subtle behavioral shifts—such as reduced movement, altered eating habits, or abnormal social interactions—often days before clinical symptoms appear <cite class="ltx_cite ltx_citemacro_cite">[<a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#bib.bib19" title="">19</a>, <a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#bib.bib20" title="">20</a>]</cite>. These automated surveillance tools address key bottlenecks in manual disease monitoring, which is labor-intensive and susceptible to observational errors. By automating clinical sign detection and aggregating herd data, AI-based systems significantly improve the speed, accuracy, and scalability of swine disease surveillance.</p> </div> <div class="ltx_para ltx_noindent" id="S2.p7"> <p class="ltx_p" id="S2.p7.1">This study refines multi-agent AI frameworks to address swine disease detection challenges, integrating proven methodologies from human healthcare with veterinary diagnostics. By coordinating specialized agents—such as domain knowledge retrievers and diagnostic decision-makers—the system enables timely, evidence-based veterinary support. It enhances early disease identification, guides targeted interventions, and mitigates economic losses. Ultimately, this research contributes to a more resilient livestock sector, where AI-driven surveillance strengthens veterinary expertise and supports global food security.</p> </div> </section> <section class="ltx_section" id="S3"> <h2 class="ltx_title ltx_title_section"> <span class="ltx_tag ltx_tag_section">3 </span>Methodology</h2> <div class="ltx_para ltx_noindent" id="S3.p1"> <p class="ltx_p" id="S3.p1.1">The system follows a structured three-stage methodology, beginning with <span class="ltx_text ltx_font_bold" id="S3.p1.1.1">Query Classification <a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#S3.SS1" title="3.1 Query Classification ‣ 3 Methodology ‣ When Pigs Get Sick: Multi-Agent AI for Swine Disease Detection"><span class="ltx_text ltx_ref_tag">3.1</span></a></span>, which categorizes user queries to ensure appropriate processing. Next, <span class="ltx_text ltx_font_bold" id="S3.p1.1.2">Multi-Agent Disease Prediction <a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#S3.SS2" title="3.2 Multi-Agent Disease Prediction ‣ 3 Methodology ‣ When Pigs Get Sick: Multi-Agent AI for Swine Disease Detection"><span class="ltx_text ltx_ref_tag">3.2</span></a></span> utilizes specialized agents to analyze symptom-based queries, applying confidence-weighted decision fusion to identify the most probable diagnoses. Finally, <span class="ltx_text ltx_font_bold" id="S3.p1.1.3">Recommendations <a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#S3.SS3" title="3.3 Recommendations ‣ 3 Methodology ‣ When Pigs Get Sick: Multi-Agent AI for Swine Disease Detection"><span class="ltx_text ltx_ref_tag">3.3</span></a></span> leverage RAG to provide targeted veterinary guidance.</p> </div> <figure class="ltx_figure" id="S3.F1"><svg class="ltx_picture ltx_centering" height="466.32" id="S3.F1.pic1" overflow="visible" version="1.1" width="601.93"><g fill="#000000" stroke="#000000" transform="translate(0,466.32) matrix(1 0 0 -1 0 0) translate(300.97,0) translate(0,438.72)"><g stroke-width="0.4pt"><g fill="#E6E6E6"><path d="M 154.71 27.33 L -154.71 27.33 C -156.24 27.33 -157.48 26.09 -157.48 24.56 L -157.48 -24.56 C -157.48 -26.09 -156.24 -27.33 -154.71 -27.33 L 154.71 -27.33 C 156.24 -27.33 157.48 -26.09 157.48 -24.56 L 157.48 24.56 C 157.48 26.09 156.24 27.33 154.71 27.33 Z M -157.48 -27.33"></path></g><g fill="#000000" stroke="#000000" transform="matrix(1.0 0.0 0.0 1.0 -149.36 -22.72)"><g class="ltx_tikzmatrix" transform="matrix(1 0 0 -1 0 39.28)"><g class="ltx_tikzmatrix_row" transform="matrix(1 0 0 1 0 33.13)"><g class="ltx_tikzmatrix_col ltx_nopad_l ltx_nopad_r" transform="matrix(1 0 0 -1 0 0)"><g class="ltx_tikzmatrix" transform="matrix(1 0 0 -1 0 28.325)"><g class="ltx_tikzmatrix_row" transform="matrix(1 0 0 1 0 33.13)"><g class="ltx_tikzmatrix_col ltx_nopad_l ltx_nopad_r" transform="matrix(1 0 0 -1 0 0)"><foreignobject height="9.61" overflow="visible" transform="matrix(1 0 0 -1 0 16.6)" width="207.94"><span class="ltx_text ltx_font_bold" id="S3.F1.pic1.1.1.1.1.1.1.1.1.1.1.1.1">USER:</span> Hello! What can be done? </foreignobject></g></g></g></g></g><g class="ltx_tikzmatrix_row" transform="matrix(1 0 0 1 0 42.74)"><g class="ltx_tikzmatrix_col ltx_nopad_l ltx_nopad_r" transform="matrix(1 0 0 -1 0 0)"><foreignobject height="12.3" overflow="visible" transform="matrix(1 0 0 -1 0 16.6)" width="299.49"><span class="ltx_text ltx_font_bold" id="S3.F1.pic1.2.2.2.2.2.2.1.1.1">BOT:</span> Hello! How can I assist you with pig care? </foreignobject></g></g></g></g><g fill="#E6E6E6"><path d="M 167.45 -47.57 L -167.45 -47.57 C -168.98 -47.57 -170.21 -48.81 -170.21 -50.33 L -170.21 -110.61 C -170.21 -112.13 -168.98 -113.37 -167.45 -113.37 L 167.45 -113.37 C 168.98 -113.37 170.21 -112.13 170.21 -110.61 L 170.21 -50.33 C 170.21 -48.81 168.98 -47.57 167.45 -47.57 Z M -170.21 -113.37"></path></g><g fill="#000000" stroke="#000000" transform="matrix(1.0 0.0 0.0 1.0 -165.6 -108.76)"><g class="ltx_tikzmatrix" transform="matrix(1 0 0 -1 0 50.43)"><g class="ltx_tikzmatrix_row" transform="matrix(1 0 0 1 0 44.28)"><g class="ltx_tikzmatrix_col ltx_nopad_l ltx_nopad_r" transform="matrix(1 0 0 -1 0 0)"><g class="ltx_tikzmatrix" transform="matrix(1 0 0 -1 0 38.13)"><g class="ltx_tikzmatrix_row" transform="matrix(1 0 0 1 0 31.98)"><g class="ltx_tikzmatrix_col ltx_nopad_l ltx_nopad_r" transform="matrix(1 0 0 -1 0 0)"><g class="ltx_tikzmatrix" transform="matrix(1 0 0 -1 0 25.83)"><g class="ltx_tikzmatrix_row" transform="matrix(1 0 0 1 0 29.3)"><g class="ltx_tikzmatrix_col ltx_nopad_l ltx_nopad_r" transform="matrix(1 0 0 -1 0 0)"><foreignobject height="12.3" overflow="visible" transform="matrix(1 0 0 -1 0 16.6)" width="331.97"><span class="ltx_text ltx_font_bold" id="S3.F1.pic1.3.3.3.3.1.1.1.1.1.1.1.1.1.1.1">USER:</span> Many pigs received from the source have died. </foreignobject></g></g></g></g></g><g class="ltx_tikzmatrix_row" transform="matrix(1 0 0 1 0 41.59)"><g class="ltx_tikzmatrix_col ltx_nopad_l ltx_nopad_r" transform="matrix(1 0 0 -1 0 0)"><foreignobject height="12.3" overflow="visible" transform="matrix(1 0 0 -1 0 16.6)" width="201.37"><span class="ltx_text ltx_font_bold" id="S3.F1.pic1.4.4.4.4.2.2.2.2.2.1.1.1">BOT:</span> I see. Are you looking for: </foreignobject></g></g></g></g></g><g class="ltx_tikzmatrix_row" transform="matrix(1 0 0 1 0 53.89)"><g class="ltx_tikzmatrix_col ltx_nopad_l ltx_nopad_r" transform="matrix(1 0 0 -1 0 0)"><text transform="matrix(1 0 0 -1 0 0)">- Disease diagnosis? or General pig disease info? </text></g></g></g></g><g fill="#E6E6E6"><path d="M 154.71 -133.61 L -154.71 -133.61 C -156.24 -133.61 -157.48 -134.85 -157.48 -136.38 L -157.48 -322.72 C -157.48 -324.25 -156.24 -325.49 -154.71 -325.49 L 154.71 -325.49 C 156.24 -325.49 157.48 -324.25 157.48 -322.72 L 157.48 -136.38 C 157.48 -134.85 156.24 -133.61 154.71 -133.61 Z M -157.48 -325.49"></path></g><g fill="#000000" stroke="#000000" transform="matrix(1.0 0.0 0.0 1.0 -150.9 -320.88)"><g class="ltx_tikzmatrix" transform="matrix(1 0 0 -1 0 176.5)"><g class="ltx_tikzmatrix_row" transform="matrix(1 0 0 1 0 170.35)"><g class="ltx_tikzmatrix_col ltx_nopad_l ltx_nopad_r" transform="matrix(1 0 0 -1 0 0)"><g class="ltx_tikzmatrix" transform="matrix(1 0 0 -1 0 165.545)"><g class="ltx_tikzmatrix_row" transform="matrix(1 0 0 1 0 160.75)"><g class="ltx_tikzmatrix_col ltx_nopad_l ltx_nopad_r" transform="matrix(1 0 0 -1 0 0)"><g class="ltx_tikzmatrix" transform="matrix(1 0 0 -1 0 155.945)"><g class="ltx_tikzmatrix_row" transform="matrix(1 0 0 1 0 151.14)"><g class="ltx_tikzmatrix_col ltx_nopad_l ltx_nopad_r" transform="matrix(1 0 0 -1 0 0)"><g class="ltx_tikzmatrix" transform="matrix(1 0 0 -1 0 144.99)"><g class="ltx_tikzmatrix_row" transform="matrix(1 0 0 1 0 138.84)"><g class="ltx_tikzmatrix_col ltx_nopad_l ltx_nopad_r" transform="matrix(1 0 0 -1 0 0)"><g class="ltx_tikzmatrix" transform="matrix(1 0 0 -1 0 132.69)"><g class="ltx_tikzmatrix_row" transform="matrix(1 0 0 1 0 126.54)"><g class="ltx_tikzmatrix_col ltx_nopad_l ltx_nopad_r" transform="matrix(1 0 0 -1 0 0)"><g class="ltx_tikzmatrix" transform="matrix(1 0 0 -1 0 120.39)"><g class="ltx_tikzmatrix_row" transform="matrix(1 0 0 1 0 114.24)"><g class="ltx_tikzmatrix_col ltx_nopad_l ltx_nopad_r" transform="matrix(1 0 0 -1 0 0)"><g class="ltx_tikzmatrix" transform="matrix(1 0 0 -1 0 109.435)"><g class="ltx_tikzmatrix_row" transform="matrix(1 0 0 1 0 104.63)"><g class="ltx_tikzmatrix_col ltx_nopad_l ltx_nopad_r" transform="matrix(1 0 0 -1 0 0)"><g class="ltx_tikzmatrix" transform="matrix(1 0 0 -1 0 98.48)"><g class="ltx_tikzmatrix_row" transform="matrix(1 0 0 1 0 92.33)"><g class="ltx_tikzmatrix_col ltx_nopad_l ltx_nopad_r" transform="matrix(1 0 0 -1 0 0)"><g class="ltx_tikzmatrix" transform="matrix(1 0 0 -1 0 86.18)"><g class="ltx_tikzmatrix_row" transform="matrix(1 0 0 1 0 80.03)"><g class="ltx_tikzmatrix_col ltx_nopad_l ltx_nopad_r" transform="matrix(1 0 0 -1 0 0)"><g class="ltx_tikzmatrix" transform="matrix(1 0 0 -1 0 75.225)"><g class="ltx_tikzmatrix_row" transform="matrix(1 0 0 1 0 70.42)"><g class="ltx_tikzmatrix_col ltx_nopad_l ltx_nopad_r" transform="matrix(1 0 0 -1 0 0)"><g class="ltx_tikzmatrix" transform="matrix(1 0 0 -1 0 63.5)"><g class="ltx_tikzmatrix_row" transform="matrix(1 0 0 1 0 56.58)"><g class="ltx_tikzmatrix_col ltx_nopad_l ltx_nopad_r" transform="matrix(1 0 0 -1 0 0)"><g class="ltx_tikzmatrix" transform="matrix(1 0 0 -1 0 50.43)"><g class="ltx_tikzmatrix_row" transform="matrix(1 0 0 1 0 44.28)"><g class="ltx_tikzmatrix_col ltx_nopad_l ltx_nopad_r" transform="matrix(1 0 0 -1 0 0)"><g class="ltx_tikzmatrix" transform="matrix(1 0 0 -1 0 38.13)"><g class="ltx_tikzmatrix_row" transform="matrix(1 0 0 1 0 31.98)"><g class="ltx_tikzmatrix_col ltx_nopad_l ltx_nopad_r" transform="matrix(1 0 0 -1 0 0)"><g class="ltx_tikzmatrix" transform="matrix(1 0 0 -1 0 25.83)"><g class="ltx_tikzmatrix_row" transform="matrix(1 0 0 1 0 29.3)"><g class="ltx_tikzmatrix_col ltx_nopad_l ltx_nopad_r" transform="matrix(1 0 0 -1 0 0)"><foreignobject height="12.3" overflow="visible" transform="matrix(1 0 0 -1 0 16.6)" width="262.29"><span class="ltx_text ltx_font_bold" id="S3.F1.pic1.5.5.5.5.1.1.1.1.1.1.1.1.1.1.1.1.1.1.1.1.1.1.1.1.1.1.1.1.1.1.1.1.1.1.1.1.1.1.1.1.1.1.1.1.1.1.1.1">USER:</span> Pigs have red bodies, purple ears.., </foreignobject></g></g></g></g></g><g class="ltx_tikzmatrix_row" transform="matrix(1 0 0 1 0 41.59)"><g class="ltx_tikzmatrix_col ltx_nopad_l ltx_nopad_r" transform="matrix(1 0 0 -1 0 0)"><foreignobject height="12.3" overflow="visible" transform="matrix(1 0 0 -1 0 16.6)" width="301.8"><span class="ltx_text ltx_font_bold" id="S3.F1.pic1.6.6.6.6.2.2.2.2.2.2.2.2.2.2.2.2.2.2.2.2.2.2.2.2.2.2.2.2.2.2.2.2.2.2.2.2.2.2.2.2.2.2.1.1.1">BOT:</span> Interesting. Can you provide more details? </foreignobject></g></g></g></g></g><g class="ltx_tikzmatrix_row" transform="matrix(1 0 0 1 0 53.89)"><g class="ltx_tikzmatrix_col ltx_nopad_l ltx_nopad_r" transform="matrix(1 0 0 -1 0 0)"><text transform="matrix(1 0 0 -1 0 0)">- How many pigs are affected? </text></g></g></g></g></g><g class="ltx_tikzmatrix_row" transform="matrix(1 0 0 1 0 66.96)"><g class="ltx_tikzmatrix_col ltx_nopad_l ltx_nopad_r" transform="matrix(1 0 0 -1 0 0)"><text transform="matrix(1 0 0 -1 0 0)">- What is their age/type? </text></g></g></g></g></g><g class="ltx_tikzmatrix_row" transform="matrix(1 0 0 1 0 80.03)"><g class="ltx_tikzmatrix_col ltx_nopad_l ltx_nopad_r" transform="matrix(1 0 0 -1 0 0)"><foreignobject height="9.61" overflow="visible" transform="matrix(1 0 0 -1 0 16.6)" width="253.83"><span class="ltx_text ltx_font_bold" id="S3.F1.pic1.7.7.7.7.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.3.1.1.1">USER:</span> No extra information is available. </foreignobject></g></g></g></g></g><g class="ltx_tikzmatrix_row" transform="matrix(1 0 0 1 0 89.64)"><g class="ltx_tikzmatrix_col ltx_nopad_l ltx_nopad_r" transform="matrix(1 0 0 -1 0 0)"><foreignobject height="12.3" overflow="visible" transform="matrix(1 0 0 -1 0 16.6)" width="184.07"><span class="ltx_text ltx_font_bold" id="S3.F1.pic1.8.8.8.8.4.4.4.4.4.4.4.4.4.4.4.4.4.4.4.4.4.4.4.4.4.4.4.4.4.4.1.1.1">BOT:</span> Understood. Summary: </foreignobject></g></g></g></g></g><g class="ltx_tikzmatrix_row" transform="matrix(1 0 0 1 0 101.94)"><g class="ltx_tikzmatrix_col ltx_nopad_l ltx_nopad_r" transform="matrix(1 0 0 -1 0 0)"><text transform="matrix(1 0 0 -1 0 0)">- Respiratory symptoms, Red body, purple ears, </text></g></g></g></g></g><g class="ltx_tikzmatrix_row" transform="matrix(1 0 0 1 0 114.24)"><g class="ltx_tikzmatrix_col ltx_nopad_l ltx_nopad_r" transform="matrix(1 0 0 -1 0 0)"><text transform="matrix(1 0 0 -1 0 0)">… Is this correct? </text></g></g></g></g></g><g class="ltx_tikzmatrix_row" transform="matrix(1 0 0 1 0 123.85)"><g class="ltx_tikzmatrix_col ltx_nopad_l ltx_nopad_r" transform="matrix(1 0 0 -1 0 0)"><foreignobject height="12.3" overflow="visible" transform="matrix(1 0 0 -1 0 16.6)" width="173.92"><span class="ltx_text ltx_font_bold" id="S3.F1.pic1.9.9.9.9.5.5.5.5.5.5.5.5.5.5.5.5.5.5.5.5.5.1.1.1">USER:</span> Yes, that’s accurate. </foreignobject></g></g></g></g></g><g class="ltx_tikzmatrix_row" transform="matrix(1 0 0 1 0 136.15)"><g class="ltx_tikzmatrix_col ltx_nopad_l ltx_nopad_r" transform="matrix(1 0 0 -1 0 0)"><foreignobject height="12.3" overflow="visible" transform="matrix(1 0 0 -1 0 16.6)" width="192.41"><span class="ltx_text ltx_font_bold" id="S3.F1.pic1.10.10.10.10.6.6.6.6.6.6.6.6.6.6.6.6.6.6.1.1.1">BOT:</span> Based on the symptoms: </foreignobject></g></g></g></g></g><g class="ltx_tikzmatrix_row" transform="matrix(1 0 0 1 0 148.45)"><g class="ltx_tikzmatrix_col ltx_nopad_l ltx_nopad_r" transform="matrix(1 0 0 -1 0 0)"><text transform="matrix(1 0 0 -1 0 0)">- PRRS, ASF, PED, and FMD are unlikely. </text></g></g></g></g></g><g class="ltx_tikzmatrix_row" transform="matrix(1 0 0 1 0 160.75)"><g class="ltx_tikzmatrix_col ltx_nopad_l ltx_nopad_r" transform="matrix(1 0 0 -1 0 0)"><text transform="matrix(1 0 0 -1 0 0)">- There is a small chance of ASF. </text></g></g></g></g></g><g class="ltx_tikzmatrix_row" transform="matrix(1 0 0 1 0 170.36)"><g class="ltx_tikzmatrix_col ltx_nopad_l ltx_nopad_r" transform="matrix(1 0 0 -1 0 0)"><text transform="matrix(1 0 0 -1 0 0)">Consult a vet. </text></g></g></g></g></g><g class="ltx_tikzmatrix_row" transform="matrix(1 0 0 1 0 179.96)"><g class="ltx_tikzmatrix_col ltx_nopad_l ltx_nopad_r" transform="matrix(1 0 0 -1 0 0)"><text transform="matrix(1 0 0 -1 0 0)">Would you like details on ASF testing? </text></g></g></g></g><g fill="#E6E6E6"><path d="M 154.71 -345.73 L -154.71 -345.73 C -156.24 -345.73 -157.48 -346.97 -157.48 -348.5 L -157.48 -435.67 C -157.48 -437.2 -156.24 -438.44 -154.71 -438.44 L 154.71 -438.44 C 156.24 -438.44 157.48 -437.2 157.48 -435.67 L 157.48 -348.5 C 157.48 -346.97 156.24 -345.73 154.71 -345.73 Z M -157.48 -438.44"></path></g><g fill="#000000" stroke="#000000" transform="matrix(1.0 0.0 0.0 1.0 -148.08 -433.83)"><g class="ltx_tikzmatrix" transform="matrix(1 0 0 -1 0 76.57)"><g class="ltx_tikzmatrix_row" transform="matrix(1 0 0 1 0 69.65)"><g class="ltx_tikzmatrix_col ltx_nopad_l ltx_nopad_r" transform="matrix(1 0 0 -1 0 0)"><g class="ltx_tikzmatrix" transform="matrix(1 0 0 -1 0 63.115)"><g class="ltx_tikzmatrix_row" transform="matrix(1 0 0 1 0 56.58)"><g class="ltx_tikzmatrix_col ltx_nopad_l ltx_nopad_r" transform="matrix(1 0 0 -1 0 0)"><g class="ltx_tikzmatrix" transform="matrix(1 0 0 -1 0 50.43)"><g class="ltx_tikzmatrix_row" transform="matrix(1 0 0 1 0 44.28)"><g class="ltx_tikzmatrix_col ltx_nopad_l ltx_nopad_r" transform="matrix(1 0 0 -1 0 0)"><g class="ltx_tikzmatrix" transform="matrix(1 0 0 -1 0 38.13)"><g class="ltx_tikzmatrix_row" transform="matrix(1 0 0 1 0 31.98)"><g class="ltx_tikzmatrix_col ltx_nopad_l ltx_nopad_r" transform="matrix(1 0 0 -1 0 0)"><g class="ltx_tikzmatrix" transform="matrix(1 0 0 -1 0 25.83)"><g class="ltx_tikzmatrix_row" transform="matrix(1 0 0 1 0 29.3)"><g class="ltx_tikzmatrix_col ltx_nopad_l ltx_nopad_r" transform="matrix(1 0 0 -1 0 0)"><foreignobject height="12.3" overflow="visible" transform="matrix(1 0 0 -1 0 16.6)" width="296.15"><span class="ltx_text ltx_font_bold" id="S3.F1.pic1.11.11.11.11.1.1.1.1.1.1.1.1.1.1.1.1.1.1.1.1.1">USER:</span> What samples are used for ASF testing? </foreignobject></g></g></g></g></g><g class="ltx_tikzmatrix_row" transform="matrix(1 0 0 1 0 41.59)"><g class="ltx_tikzmatrix_col ltx_nopad_l ltx_nopad_r" transform="matrix(1 0 0 -1 0 0)"><foreignobject height="12.3" overflow="visible" transform="matrix(1 0 0 -1 0 16.6)" width="191.72"><span class="ltx_text ltx_font_bold" id="S3.F1.pic1.12.12.12.12.2.2.2.2.2.2.2.2.2.2.2.1.1.1">BOT:</span> For ASF testing, collect: </foreignobject></g></g></g></g></g><g class="ltx_tikzmatrix_row" transform="matrix(1 0 0 1 0 53.89)"><g class="ltx_tikzmatrix_col ltx_nopad_l ltx_nopad_r" transform="matrix(1 0 0 -1 0 0)"><text transform="matrix(1 0 0 -1 0 0)">- Blood, saliva, lymph nodes, organ samples </text></g></g></g></g></g><g class="ltx_tikzmatrix_row" transform="matrix(1 0 0 1 0 66.96)"><g class="ltx_tikzmatrix_col ltx_nopad_l ltx_nopad_r" transform="matrix(1 0 0 -1 0 0)"><text transform="matrix(1 0 0 -1 0 0)">**Do NOT perform on-farm necropsy!** </text></g></g></g></g></g><g class="ltx_tikzmatrix_row" transform="matrix(1 0 0 1 0 80.03)"><g class="ltx_tikzmatrix_col ltx_nopad_l ltx_nopad_r" transform="matrix(1 0 0 -1 0 0)"><text transform="matrix(1 0 0 -1 0 0)">Would you like to read more? (ASF-2022.pdf) </text></g></g></g></g><g stroke-width="0.8pt"><path d="M 0 -27.6 L 0 -43.69" style="fill:none"></path><g transform="matrix(0.0 -1.0 1.0 0.0 0 -43.69)"><path d="M 3.6 0 L -2.16 2.88 L 0 0 L -2.16 -2.88" style="stroke:none"></path></g></g><g stroke-width="0.8pt"><path d="M 0 -113.65 L 0 -129.74" style="fill:none"></path><g transform="matrix(0.0 -1.0 1.0 0.0 0 -129.74)"><path d="M 3.6 0 L -2.16 2.88 L 0 0 L -2.16 -2.88" style="stroke:none"></path></g></g><g stroke-width="0.8pt"><path d="M 0 -325.77 L 0 -341.85" style="fill:none"></path><g transform="matrix(0.0 -1.0 1.0 0.0 0 -341.85)"><path d="M 3.6 0 L -2.16 2.88 L 0 0 L -2.16 -2.88" style="stroke:none"></path></g></g><g stroke-dasharray="3.0pt,3.0pt" stroke-dashoffset="0.0pt"><path d="M -193.22 -55.14 L -297.92 -55.14 C -299.45 -55.14 -300.69 -56.38 -300.69 -57.91 L -300.69 -103.03 C -300.69 -104.56 -299.45 -105.8 -297.92 -105.8 L -193.22 -105.8 C -191.69 -105.8 -190.45 -104.56 -190.45 -103.03 L -190.45 -57.91 C -190.45 -56.38 -191.69 -55.14 -193.22 -55.14 Z M -300.69 -105.8" style="fill:none"></path></g><g fill="#000000" stroke="#000000" stroke-dasharray="3.0pt,3.0pt" stroke-dashoffset="0.0pt" transform="matrix(1.0 0.0 0.0 1.0 -285.81 -101.19)"><g class="ltx_tikzmatrix" transform="matrix(1 0 0 -1 0 36.635)"><g class="ltx_tikzmatrix_row" transform="matrix(1 0 0 1 0 31.83)"><g class="ltx_tikzmatrix_col ltx_nopad_l ltx_nopad_r" transform="matrix(1 0 0 -1 19.28 0)"><g class="ltx_tikzmatrix" transform="matrix(1 0 0 -1 0 25.755)"><g class="ltx_tikzmatrix_row" transform="matrix(1 0 0 1 0 29.15)"><g class="ltx_tikzmatrix_col ltx_nopad_l ltx_nopad_r" transform="matrix(1 0 0 -1 0 0)"><text transform="matrix(1 0 0 -1 0 0)">Query </text></g></g></g></g></g><g class="ltx_tikzmatrix_row" transform="matrix(1 0 0 1 0 41.44)"><g class="ltx_tikzmatrix_col ltx_nopad_l ltx_nopad_r" transform="matrix(1 0 0 -1 0 0)"><text transform="matrix(1 0 0 -1 0 0)">Classification</text></g></g></g></g><g stroke-dasharray="3.0pt,3.0pt" stroke-dashoffset="0.0pt"><path d="M -180.49 -199.41 L -285.19 -199.41 C -286.72 -199.41 -287.96 -200.65 -287.96 -202.18 L -287.96 -256.92 C -287.96 -258.45 -286.72 -259.69 -285.19 -259.69 L -180.49 -259.69 C -178.96 -259.69 -177.72 -258.45 -177.72 -256.92 L -177.72 -202.18 C -177.72 -200.65 -178.96 -199.41 -180.49 -199.41 Z M -287.96 -259.69" style="fill:none"></path></g><g fill="#000000" stroke="#000000" stroke-dasharray="3.0pt,3.0pt" stroke-dashoffset="0.0pt" transform="matrix(1.0 0.0 0.0 1.0 -272.23 -255.08)"><g class="ltx_tikzmatrix" transform="matrix(1 0 0 -1 0 46.245)"><g class="ltx_tikzmatrix_row" transform="matrix(1 0 0 1 0 41.44)"><g class="ltx_tikzmatrix_col ltx_nopad_l ltx_nopad_r" transform="matrix(1 0 0 -1 0 0)"><g class="ltx_tikzmatrix" transform="matrix(1 0 0 -1 0 36.71)"><g class="ltx_tikzmatrix_row" transform="matrix(1 0 0 1 0 31.98)"><g class="ltx_tikzmatrix_col ltx_nopad_l ltx_nopad_r" transform="matrix(1 0 0 -1 0 0)"><g class="ltx_tikzmatrix" transform="matrix(1 0 0 -1 0 25.83)"><g class="ltx_tikzmatrix_row" transform="matrix(1 0 0 1 0 29.3)"><g class="ltx_tikzmatrix_col ltx_nopad_l ltx_nopad_r" transform="matrix(1 0 0 -1 0 0)"><text transform="matrix(1 0 0 -1 0 0)">Multi-Agent </text></g></g></g></g></g><g class="ltx_tikzmatrix_row" transform="matrix(1 0 0 1 0 41.44)"><g class="ltx_tikzmatrix_col ltx_nopad_l ltx_nopad_r" transform="matrix(1 0 0 -1 14.82 0)"><text transform="matrix(1 0 0 -1 0 0)">Disease </text></g></g></g></g></g><g class="ltx_tikzmatrix_row" transform="matrix(1 0 0 1 0 51.05)"><g class="ltx_tikzmatrix_col ltx_nopad_l ltx_nopad_r" transform="matrix(1 0 0 -1 8.15 0)"><text transform="matrix(1 0 0 -1 0 0)">Prediction</text></g></g></g></g><g stroke-dasharray="3.0pt,3.0pt" stroke-dashoffset="0.0pt"><path d="M -180.49 -372.83 L -294.37 -372.83 C -295.9 -372.83 -297.14 -374.06 -297.14 -375.59 L -297.14 -408.58 C -297.14 -410.1 -295.9 -411.34 -294.37 -411.34 L -180.49 -411.34 C -178.96 -411.34 -177.72 -410.1 -177.72 -408.58 L -177.72 -375.59 C -177.72 -374.06 -178.96 -372.83 -180.49 -372.83 Z M -297.14 -411.34" style="fill:none"></path></g><g fill="#000000" stroke="#000000" stroke-dasharray="3.0pt,3.0pt" stroke-dashoffset="0.0pt" transform="matrix(1.0 0.0 0.0 1.0 -292.53 -406.73)"><g class="ltx_tikzmatrix" transform="matrix(1 0 0 -1 0 24.485)"><g class="ltx_tikzmatrix_row" transform="matrix(1 0 0 1 0 29.3)"><g class="ltx_tikzmatrix_col ltx_nopad_l ltx_nopad_r" transform="matrix(1 0 0 -1 0 0)"><text transform="matrix(1 0 0 -1 0 0)">Recommendations</text></g></g></g></g><g stroke-width="0.8pt"><path d="M -245.57 -106.08 L -233.32 -195.57" style="fill:none"></path><g transform="matrix(0.13557 -0.99077 0.99077 0.13557 -233.32 -195.57)"><path d="M 3.6 0 L -2.16 2.88 L 0 0 L -2.16 -2.88" style="stroke:none"></path></g></g><g stroke-width="0.8pt"><path d="M -232.84 -259.96 L -237.28 -368.95" style="fill:none"></path><g transform="matrix(-0.04076 -0.99918 0.99918 -0.04076 -237.28 -368.95)"><path d="M 3.6 0 L -2.16 2.88 L 0 0 L -2.16 -2.88" style="stroke:none"></path></g></g><g stroke-dasharray="3.0pt,3.0pt" stroke-dashoffset="0.0pt"><path d="M 285.19 19.26 L 180.49 19.26 C 178.96 19.26 177.72 18.02 177.72 16.49 L 177.72 -16.49 C 177.72 -18.02 178.96 -19.26 180.49 -19.26 L 285.19 -19.26 C 286.72 -19.26 287.96 -18.02 287.96 -16.49 L 287.96 16.49 C 287.96 18.02 286.72 19.26 285.19 19.26 Z M 177.72 -19.26" style="fill:none"></path></g><g fill="#000000" stroke="#000000" stroke-dasharray="3.0pt,3.0pt" stroke-dashoffset="0.0pt" transform="matrix(1.0 0.0 0.0 1.0 209.32 -14.65)"><g class="ltx_tikzmatrix" transform="matrix(1 0 0 -1 0 24.485)"><g class="ltx_tikzmatrix_row" transform="matrix(1 0 0 1 0 29.3)"><g class="ltx_tikzmatrix_col ltx_nopad_l ltx_nopad_r" transform="matrix(1 0 0 -1 0 0)"><text transform="matrix(1 0 0 -1 0 0)">General</text></g></g></g></g><g stroke-dasharray="3.0pt,3.0pt" stroke-dashoffset="0.0pt"><path d="M 297.92 -61.29 L 193.22 -61.29 C 191.69 -61.29 190.45 -62.53 190.45 -64.06 L 190.45 -96.88 C 190.45 -98.41 191.69 -99.65 193.22 -99.65 L 297.92 -99.65 C 299.45 -99.65 300.69 -98.41 300.69 -96.88 L 300.69 -64.06 C 300.69 -62.53 299.45 -61.29 297.92 -61.29 Z M 190.45 -99.65" style="fill:none"></path></g><g fill="#000000" stroke="#000000" stroke-dasharray="3.0pt,3.0pt" stroke-dashoffset="0.0pt" transform="matrix(1.0 0.0 0.0 1.0 230.68 -95.04)"><g class="ltx_tikzmatrix" transform="matrix(1 0 0 -1 0 24.41)"><g class="ltx_tikzmatrix_row" transform="matrix(1 0 0 1 0 29.15)"><g class="ltx_tikzmatrix_col ltx_nopad_l ltx_nopad_r" transform="matrix(1 0 0 -1 0 0)"><text transform="matrix(1 0 0 -1 0 0)">TBC</text></g></g></g></g><g stroke-dasharray="3.0pt,3.0pt" stroke-dashoffset="0.0pt"><path d="M 285.19 -209.02 L 180.49 -209.02 C 178.96 -209.02 177.72 -210.26 177.72 -211.79 L 177.72 -247.31 C 177.72 -248.84 178.96 -250.08 180.49 -250.08 L 285.19 -250.08 C 286.72 -250.08 287.96 -248.84 287.96 -247.31 L 287.96 -211.79 C 287.96 -210.26 286.72 -209.02 285.19 -209.02 Z M 177.72 -250.08" style="fill:none"></path></g><g fill="#000000" stroke="#000000" stroke-dasharray="3.0pt,3.0pt" stroke-dashoffset="0.0pt" transform="matrix(1.0 0.0 0.0 1.0 200.99 -245.47)"><g class="ltx_tikzmatrix" transform="matrix(1 0 0 -1 0 25.755)"><g class="ltx_tikzmatrix_row" transform="matrix(1 0 0 1 0 29.15)"><g class="ltx_tikzmatrix_col ltx_nopad_l ltx_nopad_r" transform="matrix(1 0 0 -1 0 0)"><text transform="matrix(1 0 0 -1 0 0)">Diagnostic</text></g></g></g></g><g stroke-dasharray="3.0pt,3.0pt" stroke-dashoffset="0.0pt"><path d="M 285.19 -366.68 L 180.49 -366.68 C 178.96 -366.68 177.72 -367.92 177.72 -369.44 L 177.72 -414.73 C 177.72 -416.25 178.96 -417.49 180.49 -417.49 L 285.19 -417.49 C 286.72 -417.49 287.96 -416.25 287.96 -414.73 L 287.96 -369.44 C 287.96 -367.92 286.72 -366.68 285.19 -366.68 Z M 177.72 -417.49" style="fill:none"></path></g><g fill="#000000" stroke="#000000" stroke-dasharray="3.0pt,3.0pt" stroke-dashoffset="0.0pt" transform="matrix(1.0 0.0 0.0 1.0 197.67 -412.88)"><g class="ltx_tikzmatrix" transform="matrix(1 0 0 -1 0 36.785)"><g class="ltx_tikzmatrix_row" transform="matrix(1 0 0 1 0 31.98)"><g class="ltx_tikzmatrix_col ltx_nopad_l ltx_nopad_r" transform="matrix(1 0 0 -1 0 0)"><g class="ltx_tikzmatrix" transform="matrix(1 0 0 -1 0 25.83)"><g class="ltx_tikzmatrix_row" transform="matrix(1 0 0 1 0 29.3)"><g class="ltx_tikzmatrix_col ltx_nopad_l ltx_nopad_r" transform="matrix(1 0 0 -1 0 0)"><text transform="matrix(1 0 0 -1 0 0)">Knowledge </text></g></g></g></g></g><g class="ltx_tikzmatrix_row" transform="matrix(1 0 0 1 0 41.59)"><g class="ltx_tikzmatrix_col ltx_nopad_l ltx_nopad_r" transform="matrix(1 0 0 -1 7.96 0)"><text transform="matrix(1 0 0 -1 0 0)">Retrieval</text></g></g></g></g></g><g stroke-width="0.8pt"><path d="M 232.84 -19.54 L 244.51 -57.57" style="fill:none"></path><g transform="matrix(0.2935 -0.95596 0.95596 0.2935 244.51 -57.57)"><path d="M 3.6 0 L -2.16 2.88 L 0 0 L -2.16 -2.88" style="stroke:none"></path></g></g><g stroke-width="0.8pt"><path d="M 245.57 -99.93 L 233.26 -205.17" style="fill:none"></path><g transform="matrix(-0.11623 -0.99323 0.99323 -0.11623 233.26 -205.17)"><path d="M 3.6 0 L -2.16 2.88 L 0 0 L -2.16 -2.88" style="stroke:none"></path></g></g><g stroke-width="0.8pt"><path d="M 232.84 -250.36 L 232.84 -362.8" style="fill:none"></path><g transform="matrix(0.0 -1.0 1.0 0.0 232.84 -362.8)"><path d="M 3.6 0 L -2.16 2.88 L 0 0 L -2.16 -2.88" style="stroke:none"></path></g></g></g></svg> <figcaption class="ltx_caption ltx_centering"><span class="ltx_tag ltx_tag_figure">Figure 1: </span>Schematic representation of the multi-turn diagnostic conversation flow. This diagram illustrates how user queries transition from initial query classification to subsequent processing by specialized modules— including multi-agent disease prediction and RAG—ultimately leading to actionable veterinary recommendations.</figcaption> </figure> <section class="ltx_subsection" id="S3.SS1"> <h3 class="ltx_title ltx_title_subsection"> <span class="ltx_tag ltx_tag_subsection">3.1 </span>Query Classification</h3> <div class="ltx_para ltx_noindent" id="S3.SS1.p1"> <p class="ltx_p" id="S3.SS1.p1.4">To enhance diagnostic accuracy and response efficiency, user queries are first categorized into four classes:</p> <table class="ltx_equation ltx_eqn_table" id="S3.Ex1"> <tbody><tr class="ltx_equation ltx_eqn_row ltx_align_baseline"> <td class="ltx_eqn_cell ltx_eqn_center_padleft"></td> <td class="ltx_eqn_cell ltx_align_center"><math alttext="\mathcal{C}\;=\;\{\text{K},\;\text{D},\;\text{T},\;\text{G}\}." class="ltx_Math" display="block" id="S3.Ex1.m1.5"><semantics id="S3.Ex1.m1.5a"><mrow id="S3.Ex1.m1.5.5.1" xref="S3.Ex1.m1.5.5.1.1.cmml"><mrow id="S3.Ex1.m1.5.5.1.1" xref="S3.Ex1.m1.5.5.1.1.cmml"><mi class="ltx_font_mathcaligraphic" id="S3.Ex1.m1.5.5.1.1.2" xref="S3.Ex1.m1.5.5.1.1.2.cmml">𝒞</mi><mo id="S3.Ex1.m1.5.5.1.1.1" lspace="0.558em" rspace="0.558em" xref="S3.Ex1.m1.5.5.1.1.1.cmml">=</mo><mrow id="S3.Ex1.m1.5.5.1.1.3.2" xref="S3.Ex1.m1.5.5.1.1.3.1.cmml"><mo id="S3.Ex1.m1.5.5.1.1.3.2.1" stretchy="false" xref="S3.Ex1.m1.5.5.1.1.3.1.cmml">{</mo><mtext id="S3.Ex1.m1.1.1" xref="S3.Ex1.m1.1.1a.cmml">K</mtext><mo id="S3.Ex1.m1.5.5.1.1.3.2.2" rspace="0.447em" xref="S3.Ex1.m1.5.5.1.1.3.1.cmml">,</mo><mtext id="S3.Ex1.m1.2.2" xref="S3.Ex1.m1.2.2a.cmml">D</mtext><mo id="S3.Ex1.m1.5.5.1.1.3.2.3" rspace="0.447em" xref="S3.Ex1.m1.5.5.1.1.3.1.cmml">,</mo><mtext id="S3.Ex1.m1.3.3" xref="S3.Ex1.m1.3.3a.cmml">T</mtext><mo id="S3.Ex1.m1.5.5.1.1.3.2.4" rspace="0.447em" xref="S3.Ex1.m1.5.5.1.1.3.1.cmml">,</mo><mtext id="S3.Ex1.m1.4.4" xref="S3.Ex1.m1.4.4a.cmml">G</mtext><mo id="S3.Ex1.m1.5.5.1.1.3.2.5" stretchy="false" xref="S3.Ex1.m1.5.5.1.1.3.1.cmml">}</mo></mrow></mrow><mo id="S3.Ex1.m1.5.5.1.2" lspace="0em" xref="S3.Ex1.m1.5.5.1.1.cmml">.</mo></mrow><annotation-xml encoding="MathML-Content" id="S3.Ex1.m1.5b"><apply id="S3.Ex1.m1.5.5.1.1.cmml" xref="S3.Ex1.m1.5.5.1"><eq id="S3.Ex1.m1.5.5.1.1.1.cmml" xref="S3.Ex1.m1.5.5.1.1.1"></eq><ci id="S3.Ex1.m1.5.5.1.1.2.cmml" xref="S3.Ex1.m1.5.5.1.1.2">𝒞</ci><set id="S3.Ex1.m1.5.5.1.1.3.1.cmml" xref="S3.Ex1.m1.5.5.1.1.3.2"><ci id="S3.Ex1.m1.1.1a.cmml" xref="S3.Ex1.m1.1.1"><mtext id="S3.Ex1.m1.1.1.cmml" xref="S3.Ex1.m1.1.1">K</mtext></ci><ci id="S3.Ex1.m1.2.2a.cmml" xref="S3.Ex1.m1.2.2"><mtext id="S3.Ex1.m1.2.2.cmml" xref="S3.Ex1.m1.2.2">D</mtext></ci><ci id="S3.Ex1.m1.3.3a.cmml" xref="S3.Ex1.m1.3.3"><mtext id="S3.Ex1.m1.3.3.cmml" xref="S3.Ex1.m1.3.3">T</mtext></ci><ci id="S3.Ex1.m1.4.4a.cmml" xref="S3.Ex1.m1.4.4"><mtext id="S3.Ex1.m1.4.4.cmml" xref="S3.Ex1.m1.4.4">G</mtext></ci></set></apply></annotation-xml><annotation encoding="application/x-tex" id="S3.Ex1.m1.5c">\mathcal{C}\;=\;\{\text{K},\;\text{D},\;\text{T},\;\text{G}\}.</annotation><annotation encoding="application/x-llamapun" id="S3.Ex1.m1.5d">caligraphic_C = { K , D , T , G } .</annotation></semantics></math></td> <td class="ltx_eqn_cell ltx_eqn_center_padright"></td> </tr></tbody> </table> <p class="ltx_p" id="S3.SS1.p1.3">Given a user query <math alttext="q" class="ltx_Math" display="inline" id="S3.SS1.p1.1.m1.1"><semantics id="S3.SS1.p1.1.m1.1a"><mi id="S3.SS1.p1.1.m1.1.1" xref="S3.SS1.p1.1.m1.1.1.cmml">q</mi><annotation-xml encoding="MathML-Content" id="S3.SS1.p1.1.m1.1b"><ci id="S3.SS1.p1.1.m1.1.1.cmml" xref="S3.SS1.p1.1.m1.1.1">𝑞</ci></annotation-xml><annotation encoding="application/x-tex" id="S3.SS1.p1.1.m1.1c">q</annotation><annotation encoding="application/x-llamapun" id="S3.SS1.p1.1.m1.1d">italic_q</annotation></semantics></math>, a classification model provides probabilities <math alttext="P(c\mid q)" class="ltx_Math" display="inline" id="S3.SS1.p1.2.m2.1"><semantics id="S3.SS1.p1.2.m2.1a"><mrow id="S3.SS1.p1.2.m2.1.1" xref="S3.SS1.p1.2.m2.1.1.cmml"><mi id="S3.SS1.p1.2.m2.1.1.3" xref="S3.SS1.p1.2.m2.1.1.3.cmml">P</mi><mo id="S3.SS1.p1.2.m2.1.1.2" xref="S3.SS1.p1.2.m2.1.1.2.cmml">⁢</mo><mrow id="S3.SS1.p1.2.m2.1.1.1.1" xref="S3.SS1.p1.2.m2.1.1.1.1.1.cmml"><mo id="S3.SS1.p1.2.m2.1.1.1.1.2" stretchy="false" xref="S3.SS1.p1.2.m2.1.1.1.1.1.cmml">(</mo><mrow id="S3.SS1.p1.2.m2.1.1.1.1.1" xref="S3.SS1.p1.2.m2.1.1.1.1.1.cmml"><mi id="S3.SS1.p1.2.m2.1.1.1.1.1.2" xref="S3.SS1.p1.2.m2.1.1.1.1.1.2.cmml">c</mi><mo id="S3.SS1.p1.2.m2.1.1.1.1.1.1" xref="S3.SS1.p1.2.m2.1.1.1.1.1.1.cmml">∣</mo><mi id="S3.SS1.p1.2.m2.1.1.1.1.1.3" xref="S3.SS1.p1.2.m2.1.1.1.1.1.3.cmml">q</mi></mrow><mo id="S3.SS1.p1.2.m2.1.1.1.1.3" stretchy="false" xref="S3.SS1.p1.2.m2.1.1.1.1.1.cmml">)</mo></mrow></mrow><annotation-xml encoding="MathML-Content" id="S3.SS1.p1.2.m2.1b"><apply id="S3.SS1.p1.2.m2.1.1.cmml" xref="S3.SS1.p1.2.m2.1.1"><times id="S3.SS1.p1.2.m2.1.1.2.cmml" xref="S3.SS1.p1.2.m2.1.1.2"></times><ci id="S3.SS1.p1.2.m2.1.1.3.cmml" xref="S3.SS1.p1.2.m2.1.1.3">𝑃</ci><apply id="S3.SS1.p1.2.m2.1.1.1.1.1.cmml" xref="S3.SS1.p1.2.m2.1.1.1.1"><csymbol cd="latexml" id="S3.SS1.p1.2.m2.1.1.1.1.1.1.cmml" xref="S3.SS1.p1.2.m2.1.1.1.1.1.1">conditional</csymbol><ci id="S3.SS1.p1.2.m2.1.1.1.1.1.2.cmml" xref="S3.SS1.p1.2.m2.1.1.1.1.1.2">𝑐</ci><ci id="S3.SS1.p1.2.m2.1.1.1.1.1.3.cmml" xref="S3.SS1.p1.2.m2.1.1.1.1.1.3">𝑞</ci></apply></apply></annotation-xml><annotation encoding="application/x-tex" id="S3.SS1.p1.2.m2.1c">P(c\mid q)</annotation><annotation encoding="application/x-llamapun" id="S3.SS1.p1.2.m2.1d">italic_P ( italic_c ∣ italic_q )</annotation></semantics></math> for each <math alttext="c\in\mathcal{C}" class="ltx_Math" display="inline" id="S3.SS1.p1.3.m3.1"><semantics id="S3.SS1.p1.3.m3.1a"><mrow id="S3.SS1.p1.3.m3.1.1" xref="S3.SS1.p1.3.m3.1.1.cmml"><mi id="S3.SS1.p1.3.m3.1.1.2" xref="S3.SS1.p1.3.m3.1.1.2.cmml">c</mi><mo id="S3.SS1.p1.3.m3.1.1.1" xref="S3.SS1.p1.3.m3.1.1.1.cmml">∈</mo><mi class="ltx_font_mathcaligraphic" id="S3.SS1.p1.3.m3.1.1.3" xref="S3.SS1.p1.3.m3.1.1.3.cmml">𝒞</mi></mrow><annotation-xml encoding="MathML-Content" id="S3.SS1.p1.3.m3.1b"><apply id="S3.SS1.p1.3.m3.1.1.cmml" xref="S3.SS1.p1.3.m3.1.1"><in id="S3.SS1.p1.3.m3.1.1.1.cmml" xref="S3.SS1.p1.3.m3.1.1.1"></in><ci id="S3.SS1.p1.3.m3.1.1.2.cmml" xref="S3.SS1.p1.3.m3.1.1.2">𝑐</ci><ci id="S3.SS1.p1.3.m3.1.1.3.cmml" xref="S3.SS1.p1.3.m3.1.1.3">𝒞</ci></apply></annotation-xml><annotation encoding="application/x-tex" id="S3.SS1.p1.3.m3.1c">c\in\mathcal{C}</annotation><annotation encoding="application/x-llamapun" id="S3.SS1.p1.3.m3.1d">italic_c ∈ caligraphic_C</annotation></semantics></math>. The system then chooses</p> <table class="ltx_equation ltx_eqn_table" id="S3.Ex2"> <tbody><tr class="ltx_equation ltx_eqn_row ltx_align_baseline"> <td class="ltx_eqn_cell ltx_eqn_center_padleft"></td> <td class="ltx_eqn_cell ltx_align_center"><math alttext="\hat{c}\;=\;\arg\max_{c\,\in\,\mathcal{C}}\;P(c\mid q)," class="ltx_Math" display="block" id="S3.Ex2.m1.1"><semantics id="S3.Ex2.m1.1a"><mrow id="S3.Ex2.m1.1.1.1" xref="S3.Ex2.m1.1.1.1.1.cmml"><mrow id="S3.Ex2.m1.1.1.1.1" xref="S3.Ex2.m1.1.1.1.1.cmml"><mover accent="true" id="S3.Ex2.m1.1.1.1.1.3" xref="S3.Ex2.m1.1.1.1.1.3.cmml"><mi id="S3.Ex2.m1.1.1.1.1.3.2" xref="S3.Ex2.m1.1.1.1.1.3.2.cmml">c</mi><mo id="S3.Ex2.m1.1.1.1.1.3.1" xref="S3.Ex2.m1.1.1.1.1.3.1.cmml">^</mo></mover><mo id="S3.Ex2.m1.1.1.1.1.2" lspace="0.558em" rspace="0.558em" xref="S3.Ex2.m1.1.1.1.1.2.cmml">=</mo><mrow id="S3.Ex2.m1.1.1.1.1.1" xref="S3.Ex2.m1.1.1.1.1.1.cmml"><mrow id="S3.Ex2.m1.1.1.1.1.1.3" xref="S3.Ex2.m1.1.1.1.1.1.3.cmml"><mi id="S3.Ex2.m1.1.1.1.1.1.3.1" xref="S3.Ex2.m1.1.1.1.1.1.3.1.cmml">arg</mi><mo id="S3.Ex2.m1.1.1.1.1.1.3a" lspace="0.167em" xref="S3.Ex2.m1.1.1.1.1.1.3.cmml">⁡</mo><mrow id="S3.Ex2.m1.1.1.1.1.1.3.2" xref="S3.Ex2.m1.1.1.1.1.1.3.2.cmml"><munder id="S3.Ex2.m1.1.1.1.1.1.3.2.1" xref="S3.Ex2.m1.1.1.1.1.1.3.2.1.cmml"><mi id="S3.Ex2.m1.1.1.1.1.1.3.2.1.2" xref="S3.Ex2.m1.1.1.1.1.1.3.2.1.2.cmml">max</mi><mrow id="S3.Ex2.m1.1.1.1.1.1.3.2.1.3" xref="S3.Ex2.m1.1.1.1.1.1.3.2.1.3.cmml"><mi id="S3.Ex2.m1.1.1.1.1.1.3.2.1.3.2" xref="S3.Ex2.m1.1.1.1.1.1.3.2.1.3.2.cmml">c</mi><mo id="S3.Ex2.m1.1.1.1.1.1.3.2.1.3.1" lspace="0.448em" rspace="0.448em" xref="S3.Ex2.m1.1.1.1.1.1.3.2.1.3.1.cmml">∈</mo><mi class="ltx_font_mathcaligraphic" id="S3.Ex2.m1.1.1.1.1.1.3.2.1.3.3" xref="S3.Ex2.m1.1.1.1.1.1.3.2.1.3.3.cmml">𝒞</mi></mrow></munder><mo id="S3.Ex2.m1.1.1.1.1.1.3.2a" lspace="0.167em" xref="S3.Ex2.m1.1.1.1.1.1.3.2.cmml">⁡</mo><mi id="S3.Ex2.m1.1.1.1.1.1.3.2.2" xref="S3.Ex2.m1.1.1.1.1.1.3.2.2.cmml">P</mi></mrow></mrow><mo id="S3.Ex2.m1.1.1.1.1.1.2" xref="S3.Ex2.m1.1.1.1.1.1.2.cmml">⁢</mo><mrow id="S3.Ex2.m1.1.1.1.1.1.1.1" xref="S3.Ex2.m1.1.1.1.1.1.1.1.1.cmml"><mo id="S3.Ex2.m1.1.1.1.1.1.1.1.2" stretchy="false" xref="S3.Ex2.m1.1.1.1.1.1.1.1.1.cmml">(</mo><mrow id="S3.Ex2.m1.1.1.1.1.1.1.1.1" xref="S3.Ex2.m1.1.1.1.1.1.1.1.1.cmml"><mi id="S3.Ex2.m1.1.1.1.1.1.1.1.1.2" xref="S3.Ex2.m1.1.1.1.1.1.1.1.1.2.cmml">c</mi><mo id="S3.Ex2.m1.1.1.1.1.1.1.1.1.1" xref="S3.Ex2.m1.1.1.1.1.1.1.1.1.1.cmml">∣</mo><mi id="S3.Ex2.m1.1.1.1.1.1.1.1.1.3" xref="S3.Ex2.m1.1.1.1.1.1.1.1.1.3.cmml">q</mi></mrow><mo id="S3.Ex2.m1.1.1.1.1.1.1.1.3" stretchy="false" xref="S3.Ex2.m1.1.1.1.1.1.1.1.1.cmml">)</mo></mrow></mrow></mrow><mo id="S3.Ex2.m1.1.1.1.2" xref="S3.Ex2.m1.1.1.1.1.cmml">,</mo></mrow><annotation-xml encoding="MathML-Content" id="S3.Ex2.m1.1b"><apply id="S3.Ex2.m1.1.1.1.1.cmml" xref="S3.Ex2.m1.1.1.1"><eq id="S3.Ex2.m1.1.1.1.1.2.cmml" xref="S3.Ex2.m1.1.1.1.1.2"></eq><apply id="S3.Ex2.m1.1.1.1.1.3.cmml" xref="S3.Ex2.m1.1.1.1.1.3"><ci id="S3.Ex2.m1.1.1.1.1.3.1.cmml" xref="S3.Ex2.m1.1.1.1.1.3.1">^</ci><ci id="S3.Ex2.m1.1.1.1.1.3.2.cmml" xref="S3.Ex2.m1.1.1.1.1.3.2">𝑐</ci></apply><apply id="S3.Ex2.m1.1.1.1.1.1.cmml" xref="S3.Ex2.m1.1.1.1.1.1"><times id="S3.Ex2.m1.1.1.1.1.1.2.cmml" xref="S3.Ex2.m1.1.1.1.1.1.2"></times><apply id="S3.Ex2.m1.1.1.1.1.1.3.cmml" xref="S3.Ex2.m1.1.1.1.1.1.3"><arg id="S3.Ex2.m1.1.1.1.1.1.3.1.cmml" xref="S3.Ex2.m1.1.1.1.1.1.3.1"></arg><apply id="S3.Ex2.m1.1.1.1.1.1.3.2.cmml" xref="S3.Ex2.m1.1.1.1.1.1.3.2"><apply id="S3.Ex2.m1.1.1.1.1.1.3.2.1.cmml" xref="S3.Ex2.m1.1.1.1.1.1.3.2.1"><csymbol cd="ambiguous" id="S3.Ex2.m1.1.1.1.1.1.3.2.1.1.cmml" xref="S3.Ex2.m1.1.1.1.1.1.3.2.1">subscript</csymbol><max id="S3.Ex2.m1.1.1.1.1.1.3.2.1.2.cmml" xref="S3.Ex2.m1.1.1.1.1.1.3.2.1.2"></max><apply id="S3.Ex2.m1.1.1.1.1.1.3.2.1.3.cmml" xref="S3.Ex2.m1.1.1.1.1.1.3.2.1.3"><in id="S3.Ex2.m1.1.1.1.1.1.3.2.1.3.1.cmml" xref="S3.Ex2.m1.1.1.1.1.1.3.2.1.3.1"></in><ci id="S3.Ex2.m1.1.1.1.1.1.3.2.1.3.2.cmml" xref="S3.Ex2.m1.1.1.1.1.1.3.2.1.3.2">𝑐</ci><ci id="S3.Ex2.m1.1.1.1.1.1.3.2.1.3.3.cmml" xref="S3.Ex2.m1.1.1.1.1.1.3.2.1.3.3">𝒞</ci></apply></apply><ci id="S3.Ex2.m1.1.1.1.1.1.3.2.2.cmml" xref="S3.Ex2.m1.1.1.1.1.1.3.2.2">𝑃</ci></apply></apply><apply id="S3.Ex2.m1.1.1.1.1.1.1.1.1.cmml" xref="S3.Ex2.m1.1.1.1.1.1.1.1"><csymbol cd="latexml" id="S3.Ex2.m1.1.1.1.1.1.1.1.1.1.cmml" xref="S3.Ex2.m1.1.1.1.1.1.1.1.1.1">conditional</csymbol><ci id="S3.Ex2.m1.1.1.1.1.1.1.1.1.2.cmml" xref="S3.Ex2.m1.1.1.1.1.1.1.1.1.2">𝑐</ci><ci id="S3.Ex2.m1.1.1.1.1.1.1.1.1.3.cmml" xref="S3.Ex2.m1.1.1.1.1.1.1.1.1.3">𝑞</ci></apply></apply></apply></annotation-xml><annotation encoding="application/x-tex" id="S3.Ex2.m1.1c">\hat{c}\;=\;\arg\max_{c\,\in\,\mathcal{C}}\;P(c\mid q),</annotation><annotation encoding="application/x-llamapun" id="S3.Ex2.m1.1d">over^ start_ARG italic_c end_ARG = roman_arg roman_max start_POSTSUBSCRIPT italic_c ∈ caligraphic_C end_POSTSUBSCRIPT italic_P ( italic_c ∣ italic_q ) ,</annotation></semantics></math></td> <td class="ltx_eqn_cell ltx_eqn_center_padright"></td> </tr></tbody> </table> <p class="ltx_p" id="S3.SS1.p1.5">and routes the query accordingly:</p> </div> <div class="ltx_para ltx_noindent" id="S3.SS1.p2"> <ul class="ltx_itemize" id="S3.I1"> <li class="ltx_item" id="S3.I1.i1" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item">•</span> <div class="ltx_para ltx_noindent" id="S3.I1.i1.p1"> <p class="ltx_p" id="S3.I1.i1.p1.1"><span class="ltx_text ltx_font_bold" id="S3.I1.i1.p1.1.1">K (Knowledge Retrieval Queries)</span>: These queries seek factual, evidence-based information on swine health topics, including disease prevention, vaccination protocols, prescriptions, and biosecurity measures. They correspond to the Recommendations stage <a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#S3.SS3" title="3.3 Recommendations ‣ 3 Methodology ‣ When Pigs Get Sick: Multi-Agent AI for Swine Disease Detection"><span class="ltx_text ltx_ref_tag">3.3</span></a>, meaning users can directly receive guidance without first undergoing the symptom-based diagnostic process. Instead, K queries are processed directly through RAG, enabling the system to provide immediate and actionable recommendations.</p> </div> </li> <li class="ltx_item" id="S3.I1.i2" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item">•</span> <div class="ltx_para ltx_noindent" id="S3.I1.i2.p1"> <p class="ltx_p" id="S3.I1.i2.p1.1"><span class="ltx_text ltx_font_bold" id="S3.I1.i2.p1.1.1">D (Symptom-Based Diagnostic Queries)</span>: These queries describe observed clinical signs in pigs and are analyzed by the system’s diagnostic model to generate differential diagnoses and treatment recommendations. The system collects symptom data in multiple steps before predicting the most likely diseases.</p> </div> </li> <li class="ltx_item" id="S3.I1.i3" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item">•</span> <div class="ltx_para ltx_noindent" id="S3.I1.i3.p1"> <p class="ltx_p" id="S3.I1.i3.p1.1"><span class="ltx_text ltx_font_bold" id="S3.I1.i3.p1.1.1">T (TBC Queries)</span>: TBC (to-be-clarified) queries are ambiguous, vague, or incomplete inputs. While they do not directly contribute to disease diagnostics, they prompt the user for additional details enabling accurate query classification. Effectively handling TBC queries enhances system adaptability and guides users toward more structured, actionable inputs.</p> </div> </li> <li class="ltx_item" id="S3.I1.i4" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item">•</span> <div class="ltx_para ltx_noindent" id="S3.I1.i4.p1"> <p class="ltx_p" id="S3.I1.i4.p1.1"><span class="ltx_text ltx_font_bold" id="S3.I1.i4.p1.1.1">G (General Queries)</span>: General queries include casual greetings, off-topic questions, or broad farm-related inquiries not directly related to disease. While they do not contribute to disease diagnostics, they help maintain a natural conversational flow and ensure the system can manage incomplete or irrelevant inputs.</p> </div> </li> </ul> </div> </section> <section class="ltx_subsection" id="S3.SS2"> <h3 class="ltx_title ltx_title_subsection"> <span class="ltx_tag ltx_tag_subsection">3.2 </span>Multi-Agent Disease Prediction</h3> <div class="ltx_para ltx_noindent" id="S3.SS2.p1"> <p class="ltx_p" id="S3.SS2.p1.1">Once a query is identified as a Symptom-Based Diagnostic, the system follows a structured process for disease prediction.</p> </div> <section class="ltx_subsubsection" id="S3.SS2.SSS1"> <h4 class="ltx_title ltx_title_subsubsection"> <span class="ltx_tag ltx_tag_subsubsection">3.2.1 </span>Stage-Wise Symptom Collection</h4> <div class="ltx_para ltx_noindent" id="S3.SS2.SSS1.p1"> <p class="ltx_p" id="S3.SS2.SSS1.p1.3">The system defines three key states—<span class="ltx_text ltx_font_bold" id="S3.SS2.SSS1.p1.3.1">General</span> (<math alttext="G" class="ltx_Math" display="inline" id="S3.SS2.SSS1.p1.1.m1.1"><semantics id="S3.SS2.SSS1.p1.1.m1.1a"><mi id="S3.SS2.SSS1.p1.1.m1.1.1" xref="S3.SS2.SSS1.p1.1.m1.1.1.cmml">G</mi><annotation-xml encoding="MathML-Content" id="S3.SS2.SSS1.p1.1.m1.1b"><ci id="S3.SS2.SSS1.p1.1.m1.1.1.cmml" xref="S3.SS2.SSS1.p1.1.m1.1.1">𝐺</ci></annotation-xml><annotation encoding="application/x-tex" id="S3.SS2.SSS1.p1.1.m1.1c">G</annotation><annotation encoding="application/x-llamapun" id="S3.SS2.SSS1.p1.1.m1.1d">italic_G</annotation></semantics></math>), <span class="ltx_text ltx_font_bold" id="S3.SS2.SSS1.p1.3.2">External</span> (<math alttext="E" class="ltx_Math" display="inline" id="S3.SS2.SSS1.p1.2.m2.1"><semantics id="S3.SS2.SSS1.p1.2.m2.1a"><mi id="S3.SS2.SSS1.p1.2.m2.1.1" xref="S3.SS2.SSS1.p1.2.m2.1.1.cmml">E</mi><annotation-xml encoding="MathML-Content" id="S3.SS2.SSS1.p1.2.m2.1b"><ci id="S3.SS2.SSS1.p1.2.m2.1.1.cmml" xref="S3.SS2.SSS1.p1.2.m2.1.1">𝐸</ci></annotation-xml><annotation encoding="application/x-tex" id="S3.SS2.SSS1.p1.2.m2.1c">E</annotation><annotation encoding="application/x-llamapun" id="S3.SS2.SSS1.p1.2.m2.1d">italic_E</annotation></semantics></math>), and <span class="ltx_text ltx_font_bold" id="S3.SS2.SSS1.p1.3.3">Specific</span> (<math alttext="S" class="ltx_Math" display="inline" id="S3.SS2.SSS1.p1.3.m3.1"><semantics id="S3.SS2.SSS1.p1.3.m3.1a"><mi id="S3.SS2.SSS1.p1.3.m3.1.1" xref="S3.SS2.SSS1.p1.3.m3.1.1.cmml">S</mi><annotation-xml encoding="MathML-Content" id="S3.SS2.SSS1.p1.3.m3.1b"><ci id="S3.SS2.SSS1.p1.3.m3.1.1.cmml" xref="S3.SS2.SSS1.p1.3.m3.1.1">𝑆</ci></annotation-xml><annotation encoding="application/x-tex" id="S3.SS2.SSS1.p1.3.m3.1c">S</annotation><annotation encoding="application/x-llamapun" id="S3.SS2.SSS1.p1.3.m3.1d">italic_S</annotation></semantics></math>)—and limits the symptom collection process to a maximum of three user–system exchanges before proceeding to disease prediction:</p> </div> <div class="ltx_para ltx_noindent" id="S3.SS2.SSS1.p2"> <table class="ltx_equation ltx_eqn_table" id="S3.Ex3"> <tbody><tr class="ltx_equation ltx_eqn_row ltx_align_baseline"> <td class="ltx_eqn_cell ltx_eqn_center_padleft"></td> <td class="ltx_eqn_cell ltx_align_center"><math alttext="\mathcal{S}\;=\;\{\,G,\,E,\,S\}." class="ltx_Math" display="block" id="S3.Ex3.m1.4"><semantics id="S3.Ex3.m1.4a"><mrow id="S3.Ex3.m1.4.4.1" xref="S3.Ex3.m1.4.4.1.1.cmml"><mrow id="S3.Ex3.m1.4.4.1.1" xref="S3.Ex3.m1.4.4.1.1.cmml"><mi class="ltx_font_mathcaligraphic" id="S3.Ex3.m1.4.4.1.1.2" xref="S3.Ex3.m1.4.4.1.1.2.cmml">𝒮</mi><mo id="S3.Ex3.m1.4.4.1.1.1" lspace="0.558em" rspace="0.558em" xref="S3.Ex3.m1.4.4.1.1.1.cmml">=</mo><mrow id="S3.Ex3.m1.4.4.1.1.3.2" xref="S3.Ex3.m1.4.4.1.1.3.1.cmml"><mo id="S3.Ex3.m1.4.4.1.1.3.2.1" rspace="0.170em" stretchy="false" xref="S3.Ex3.m1.4.4.1.1.3.1.cmml">{</mo><mi id="S3.Ex3.m1.1.1" xref="S3.Ex3.m1.1.1.cmml">G</mi><mo id="S3.Ex3.m1.4.4.1.1.3.2.2" rspace="0.337em" xref="S3.Ex3.m1.4.4.1.1.3.1.cmml">,</mo><mi id="S3.Ex3.m1.2.2" xref="S3.Ex3.m1.2.2.cmml">E</mi><mo id="S3.Ex3.m1.4.4.1.1.3.2.3" rspace="0.337em" xref="S3.Ex3.m1.4.4.1.1.3.1.cmml">,</mo><mi id="S3.Ex3.m1.3.3" xref="S3.Ex3.m1.3.3.cmml">S</mi><mo id="S3.Ex3.m1.4.4.1.1.3.2.4" stretchy="false" xref="S3.Ex3.m1.4.4.1.1.3.1.cmml">}</mo></mrow></mrow><mo id="S3.Ex3.m1.4.4.1.2" lspace="0em" xref="S3.Ex3.m1.4.4.1.1.cmml">.</mo></mrow><annotation-xml encoding="MathML-Content" id="S3.Ex3.m1.4b"><apply id="S3.Ex3.m1.4.4.1.1.cmml" xref="S3.Ex3.m1.4.4.1"><eq id="S3.Ex3.m1.4.4.1.1.1.cmml" xref="S3.Ex3.m1.4.4.1.1.1"></eq><ci id="S3.Ex3.m1.4.4.1.1.2.cmml" xref="S3.Ex3.m1.4.4.1.1.2">𝒮</ci><set id="S3.Ex3.m1.4.4.1.1.3.1.cmml" xref="S3.Ex3.m1.4.4.1.1.3.2"><ci id="S3.Ex3.m1.1.1.cmml" xref="S3.Ex3.m1.1.1">𝐺</ci><ci id="S3.Ex3.m1.2.2.cmml" xref="S3.Ex3.m1.2.2">𝐸</ci><ci id="S3.Ex3.m1.3.3.cmml" xref="S3.Ex3.m1.3.3">𝑆</ci></set></apply></annotation-xml><annotation encoding="application/x-tex" id="S3.Ex3.m1.4c">\mathcal{S}\;=\;\{\,G,\,E,\,S\}.</annotation><annotation encoding="application/x-llamapun" id="S3.Ex3.m1.4d">caligraphic_S = { italic_G , italic_E , italic_S } .</annotation></semantics></math></td> <td class="ltx_eqn_cell ltx_eqn_center_padright"></td> </tr></tbody> </table> <ul class="ltx_itemize" id="S3.I2"> <li class="ltx_item" id="S3.I2.i1" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item">•</span> <div class="ltx_para ltx_noindent" id="S3.I2.i1.p1"> <p class="ltx_p" id="S3.I2.i1.p1.1"><span class="ltx_text ltx_font_bold" id="S3.I2.i1.p1.1.1">State <math alttext="\boldsymbol{G}" class="ltx_Math" display="inline" id="S3.I2.i1.p1.1.1.m1.1"><semantics id="S3.I2.i1.p1.1.1.m1.1a"><mi id="S3.I2.i1.p1.1.1.m1.1.1" xref="S3.I2.i1.p1.1.1.m1.1.1.cmml">G</mi><annotation-xml encoding="MathML-Content" id="S3.I2.i1.p1.1.1.m1.1b"><ci id="S3.I2.i1.p1.1.1.m1.1.1.cmml" xref="S3.I2.i1.p1.1.1.m1.1.1">𝐺</ci></annotation-xml><annotation encoding="application/x-tex" id="S3.I2.i1.p1.1.1.m1.1c">\boldsymbol{G}</annotation><annotation encoding="application/x-llamapun" id="S3.I2.i1.p1.1.1.m1.1d">bold_italic_G</annotation></semantics></math> (General):</span> Broad health indicators such as mortality/morbidity rates, pig classification (piglets, breeders, finishers), and environmental factors (ventilation, weather changes).</p> </div> </li> <li class="ltx_item" id="S3.I2.i2" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item">•</span> <div class="ltx_para ltx_noindent" id="S3.I2.i2.p1"> <p class="ltx_p" id="S3.I2.i2.p1.1"><span class="ltx_text ltx_font_bold" id="S3.I2.i2.p1.1.1">State <math alttext="\boldsymbol{E}" class="ltx_Math" display="inline" id="S3.I2.i2.p1.1.1.m1.1"><semantics id="S3.I2.i2.p1.1.1.m1.1a"><mi id="S3.I2.i2.p1.1.1.m1.1.1" xref="S3.I2.i2.p1.1.1.m1.1.1.cmml">E</mi><annotation-xml encoding="MathML-Content" id="S3.I2.i2.p1.1.1.m1.1b"><ci id="S3.I2.i2.p1.1.1.m1.1.1.cmml" xref="S3.I2.i2.p1.1.1.m1.1.1">𝐸</ci></annotation-xml><annotation encoding="application/x-tex" id="S3.I2.i2.p1.1.1.m1.1c">\boldsymbol{E}</annotation><annotation encoding="application/x-llamapun" id="S3.I2.i2.p1.1.1.m1.1d">bold_italic_E</annotation></semantics></math> (External):</span> Visible physical signs: skin lesions, color changes, nasal or ocular discharge, and notable behavioral changes (e.g., refusal to stand, aggression).</p> </div> </li> <li class="ltx_item" id="S3.I2.i3" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item">•</span> <div class="ltx_para ltx_noindent" id="S3.I2.i3.p1"> <p class="ltx_p" id="S3.I2.i3.p1.1"><span class="ltx_text ltx_font_bold" id="S3.I2.i3.p1.1.1">State <math alttext="\boldsymbol{S}" class="ltx_Math" display="inline" id="S3.I2.i3.p1.1.1.m1.1"><semantics id="S3.I2.i3.p1.1.1.m1.1a"><mi id="S3.I2.i3.p1.1.1.m1.1.1" xref="S3.I2.i3.p1.1.1.m1.1.1.cmml">S</mi><annotation-xml encoding="MathML-Content" id="S3.I2.i3.p1.1.1.m1.1b"><ci id="S3.I2.i3.p1.1.1.m1.1.1.cmml" xref="S3.I2.i3.p1.1.1.m1.1.1">𝑆</ci></annotation-xml><annotation encoding="application/x-tex" id="S3.I2.i3.p1.1.1.m1.1c">\boldsymbol{S}</annotation><annotation encoding="application/x-llamapun" id="S3.I2.i3.p1.1.1.m1.1d">bold_italic_S</annotation></semantics></math> (Specific):</span> Targeted symptom clusters like respiratory (coughing, sneezing), gastrointestinal (vomiting, diarrhea), neurological (tremors, seizures), or reproductive (stillbirths, infertility).</p> </div> </li> </ul> </div> </section> <section class="ltx_subsubsection" id="S3.SS2.SSS2"> <h4 class="ltx_title ltx_title_subsubsection"> <span class="ltx_tag ltx_tag_subsubsection">3.2.2 </span>State Transitions</h4> <div class="ltx_para ltx_noindent" id="S3.SS2.SSS2.p1"> <p class="ltx_p" id="S3.SS2.SSS2.p1.3">At each turn <math alttext="i" class="ltx_Math" display="inline" id="S3.SS2.SSS2.p1.1.m1.1"><semantics id="S3.SS2.SSS2.p1.1.m1.1a"><mi id="S3.SS2.SSS2.p1.1.m1.1.1" xref="S3.SS2.SSS2.p1.1.m1.1.1.cmml">i</mi><annotation-xml encoding="MathML-Content" id="S3.SS2.SSS2.p1.1.m1.1b"><ci id="S3.SS2.SSS2.p1.1.m1.1.1.cmml" xref="S3.SS2.SSS2.p1.1.m1.1.1">𝑖</ci></annotation-xml><annotation encoding="application/x-tex" id="S3.SS2.SSS2.p1.1.m1.1c">i</annotation><annotation encoding="application/x-llamapun" id="S3.SS2.SSS2.p1.1.m1.1d">italic_i</annotation></semantics></math>, the system maintains a state <math alttext="s_{i}\in S" class="ltx_Math" display="inline" id="S3.SS2.SSS2.p1.2.m2.1"><semantics id="S3.SS2.SSS2.p1.2.m2.1a"><mrow id="S3.SS2.SSS2.p1.2.m2.1.1" xref="S3.SS2.SSS2.p1.2.m2.1.1.cmml"><msub id="S3.SS2.SSS2.p1.2.m2.1.1.2" xref="S3.SS2.SSS2.p1.2.m2.1.1.2.cmml"><mi id="S3.SS2.SSS2.p1.2.m2.1.1.2.2" xref="S3.SS2.SSS2.p1.2.m2.1.1.2.2.cmml">s</mi><mi id="S3.SS2.SSS2.p1.2.m2.1.1.2.3" xref="S3.SS2.SSS2.p1.2.m2.1.1.2.3.cmml">i</mi></msub><mo id="S3.SS2.SSS2.p1.2.m2.1.1.1" xref="S3.SS2.SSS2.p1.2.m2.1.1.1.cmml">∈</mo><mi id="S3.SS2.SSS2.p1.2.m2.1.1.3" xref="S3.SS2.SSS2.p1.2.m2.1.1.3.cmml">S</mi></mrow><annotation-xml encoding="MathML-Content" id="S3.SS2.SSS2.p1.2.m2.1b"><apply id="S3.SS2.SSS2.p1.2.m2.1.1.cmml" xref="S3.SS2.SSS2.p1.2.m2.1.1"><in id="S3.SS2.SSS2.p1.2.m2.1.1.1.cmml" xref="S3.SS2.SSS2.p1.2.m2.1.1.1"></in><apply id="S3.SS2.SSS2.p1.2.m2.1.1.2.cmml" xref="S3.SS2.SSS2.p1.2.m2.1.1.2"><csymbol cd="ambiguous" id="S3.SS2.SSS2.p1.2.m2.1.1.2.1.cmml" xref="S3.SS2.SSS2.p1.2.m2.1.1.2">subscript</csymbol><ci id="S3.SS2.SSS2.p1.2.m2.1.1.2.2.cmml" xref="S3.SS2.SSS2.p1.2.m2.1.1.2.2">𝑠</ci><ci id="S3.SS2.SSS2.p1.2.m2.1.1.2.3.cmml" xref="S3.SS2.SSS2.p1.2.m2.1.1.2.3">𝑖</ci></apply><ci id="S3.SS2.SSS2.p1.2.m2.1.1.3.cmml" xref="S3.SS2.SSS2.p1.2.m2.1.1.3">𝑆</ci></apply></annotation-xml><annotation encoding="application/x-tex" id="S3.SS2.SSS2.p1.2.m2.1c">s_{i}\in S</annotation><annotation encoding="application/x-llamapun" id="S3.SS2.SSS2.p1.2.m2.1d">italic_s start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT ∈ italic_S</annotation></semantics></math>, where the user’s response <math alttext="a_{i}" class="ltx_Math" display="inline" id="S3.SS2.SSS2.p1.3.m3.1"><semantics id="S3.SS2.SSS2.p1.3.m3.1a"><msub id="S3.SS2.SSS2.p1.3.m3.1.1" xref="S3.SS2.SSS2.p1.3.m3.1.1.cmml"><mi id="S3.SS2.SSS2.p1.3.m3.1.1.2" xref="S3.SS2.SSS2.p1.3.m3.1.1.2.cmml">a</mi><mi id="S3.SS2.SSS2.p1.3.m3.1.1.3" xref="S3.SS2.SSS2.p1.3.m3.1.1.3.cmml">i</mi></msub><annotation-xml encoding="MathML-Content" id="S3.SS2.SSS2.p1.3.m3.1b"><apply id="S3.SS2.SSS2.p1.3.m3.1.1.cmml" xref="S3.SS2.SSS2.p1.3.m3.1.1"><csymbol cd="ambiguous" id="S3.SS2.SSS2.p1.3.m3.1.1.1.cmml" xref="S3.SS2.SSS2.p1.3.m3.1.1">subscript</csymbol><ci id="S3.SS2.SSS2.p1.3.m3.1.1.2.cmml" xref="S3.SS2.SSS2.p1.3.m3.1.1.2">𝑎</ci><ci id="S3.SS2.SSS2.p1.3.m3.1.1.3.cmml" xref="S3.SS2.SSS2.p1.3.m3.1.1.3">𝑖</ci></apply></annotation-xml><annotation encoding="application/x-tex" id="S3.SS2.SSS2.p1.3.m3.1c">a_{i}</annotation><annotation encoding="application/x-llamapun" id="S3.SS2.SSS2.p1.3.m3.1d">italic_a start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT</annotation></semantics></math> determines the next state via the transition function:</p> <table class="ltx_equation ltx_eqn_table" id="S3.Ex4"> <tbody><tr class="ltx_equation ltx_eqn_row ltx_align_baseline"> <td class="ltx_eqn_cell ltx_eqn_center_padleft"></td> <td class="ltx_eqn_cell ltx_align_center"><math alttext="s_{i+1}\;=\;T\bigl{(}s_{i},a_{i}\bigr{)}," class="ltx_Math" display="block" id="S3.Ex4.m1.1"><semantics id="S3.Ex4.m1.1a"><mrow id="S3.Ex4.m1.1.1.1" xref="S3.Ex4.m1.1.1.1.1.cmml"><mrow id="S3.Ex4.m1.1.1.1.1" xref="S3.Ex4.m1.1.1.1.1.cmml"><msub id="S3.Ex4.m1.1.1.1.1.4" xref="S3.Ex4.m1.1.1.1.1.4.cmml"><mi id="S3.Ex4.m1.1.1.1.1.4.2" xref="S3.Ex4.m1.1.1.1.1.4.2.cmml">s</mi><mrow id="S3.Ex4.m1.1.1.1.1.4.3" xref="S3.Ex4.m1.1.1.1.1.4.3.cmml"><mi id="S3.Ex4.m1.1.1.1.1.4.3.2" xref="S3.Ex4.m1.1.1.1.1.4.3.2.cmml">i</mi><mo id="S3.Ex4.m1.1.1.1.1.4.3.1" xref="S3.Ex4.m1.1.1.1.1.4.3.1.cmml">+</mo><mn id="S3.Ex4.m1.1.1.1.1.4.3.3" xref="S3.Ex4.m1.1.1.1.1.4.3.3.cmml">1</mn></mrow></msub><mo id="S3.Ex4.m1.1.1.1.1.3" lspace="0.558em" rspace="0.558em" xref="S3.Ex4.m1.1.1.1.1.3.cmml">=</mo><mrow id="S3.Ex4.m1.1.1.1.1.2" xref="S3.Ex4.m1.1.1.1.1.2.cmml"><mi id="S3.Ex4.m1.1.1.1.1.2.4" xref="S3.Ex4.m1.1.1.1.1.2.4.cmml">T</mi><mo id="S3.Ex4.m1.1.1.1.1.2.3" xref="S3.Ex4.m1.1.1.1.1.2.3.cmml">⁢</mo><mrow id="S3.Ex4.m1.1.1.1.1.2.2.2" xref="S3.Ex4.m1.1.1.1.1.2.2.3.cmml"><mo id="S3.Ex4.m1.1.1.1.1.2.2.2.3" maxsize="120%" minsize="120%" xref="S3.Ex4.m1.1.1.1.1.2.2.3.cmml">(</mo><msub id="S3.Ex4.m1.1.1.1.1.1.1.1.1" xref="S3.Ex4.m1.1.1.1.1.1.1.1.1.cmml"><mi id="S3.Ex4.m1.1.1.1.1.1.1.1.1.2" xref="S3.Ex4.m1.1.1.1.1.1.1.1.1.2.cmml">s</mi><mi id="S3.Ex4.m1.1.1.1.1.1.1.1.1.3" xref="S3.Ex4.m1.1.1.1.1.1.1.1.1.3.cmml">i</mi></msub><mo id="S3.Ex4.m1.1.1.1.1.2.2.2.4" xref="S3.Ex4.m1.1.1.1.1.2.2.3.cmml">,</mo><msub id="S3.Ex4.m1.1.1.1.1.2.2.2.2" xref="S3.Ex4.m1.1.1.1.1.2.2.2.2.cmml"><mi id="S3.Ex4.m1.1.1.1.1.2.2.2.2.2" xref="S3.Ex4.m1.1.1.1.1.2.2.2.2.2.cmml">a</mi><mi id="S3.Ex4.m1.1.1.1.1.2.2.2.2.3" xref="S3.Ex4.m1.1.1.1.1.2.2.2.2.3.cmml">i</mi></msub><mo id="S3.Ex4.m1.1.1.1.1.2.2.2.5" maxsize="120%" minsize="120%" xref="S3.Ex4.m1.1.1.1.1.2.2.3.cmml">)</mo></mrow></mrow></mrow><mo id="S3.Ex4.m1.1.1.1.2" xref="S3.Ex4.m1.1.1.1.1.cmml">,</mo></mrow><annotation-xml encoding="MathML-Content" id="S3.Ex4.m1.1b"><apply id="S3.Ex4.m1.1.1.1.1.cmml" xref="S3.Ex4.m1.1.1.1"><eq id="S3.Ex4.m1.1.1.1.1.3.cmml" xref="S3.Ex4.m1.1.1.1.1.3"></eq><apply id="S3.Ex4.m1.1.1.1.1.4.cmml" xref="S3.Ex4.m1.1.1.1.1.4"><csymbol cd="ambiguous" id="S3.Ex4.m1.1.1.1.1.4.1.cmml" xref="S3.Ex4.m1.1.1.1.1.4">subscript</csymbol><ci id="S3.Ex4.m1.1.1.1.1.4.2.cmml" xref="S3.Ex4.m1.1.1.1.1.4.2">𝑠</ci><apply id="S3.Ex4.m1.1.1.1.1.4.3.cmml" xref="S3.Ex4.m1.1.1.1.1.4.3"><plus id="S3.Ex4.m1.1.1.1.1.4.3.1.cmml" xref="S3.Ex4.m1.1.1.1.1.4.3.1"></plus><ci id="S3.Ex4.m1.1.1.1.1.4.3.2.cmml" xref="S3.Ex4.m1.1.1.1.1.4.3.2">𝑖</ci><cn id="S3.Ex4.m1.1.1.1.1.4.3.3.cmml" type="integer" xref="S3.Ex4.m1.1.1.1.1.4.3.3">1</cn></apply></apply><apply id="S3.Ex4.m1.1.1.1.1.2.cmml" xref="S3.Ex4.m1.1.1.1.1.2"><times id="S3.Ex4.m1.1.1.1.1.2.3.cmml" xref="S3.Ex4.m1.1.1.1.1.2.3"></times><ci id="S3.Ex4.m1.1.1.1.1.2.4.cmml" xref="S3.Ex4.m1.1.1.1.1.2.4">𝑇</ci><interval closure="open" id="S3.Ex4.m1.1.1.1.1.2.2.3.cmml" xref="S3.Ex4.m1.1.1.1.1.2.2.2"><apply id="S3.Ex4.m1.1.1.1.1.1.1.1.1.cmml" xref="S3.Ex4.m1.1.1.1.1.1.1.1.1"><csymbol cd="ambiguous" id="S3.Ex4.m1.1.1.1.1.1.1.1.1.1.cmml" xref="S3.Ex4.m1.1.1.1.1.1.1.1.1">subscript</csymbol><ci id="S3.Ex4.m1.1.1.1.1.1.1.1.1.2.cmml" xref="S3.Ex4.m1.1.1.1.1.1.1.1.1.2">𝑠</ci><ci id="S3.Ex4.m1.1.1.1.1.1.1.1.1.3.cmml" xref="S3.Ex4.m1.1.1.1.1.1.1.1.1.3">𝑖</ci></apply><apply id="S3.Ex4.m1.1.1.1.1.2.2.2.2.cmml" xref="S3.Ex4.m1.1.1.1.1.2.2.2.2"><csymbol cd="ambiguous" id="S3.Ex4.m1.1.1.1.1.2.2.2.2.1.cmml" xref="S3.Ex4.m1.1.1.1.1.2.2.2.2">subscript</csymbol><ci id="S3.Ex4.m1.1.1.1.1.2.2.2.2.2.cmml" xref="S3.Ex4.m1.1.1.1.1.2.2.2.2.2">𝑎</ci><ci id="S3.Ex4.m1.1.1.1.1.2.2.2.2.3.cmml" xref="S3.Ex4.m1.1.1.1.1.2.2.2.2.3">𝑖</ci></apply></interval></apply></apply></annotation-xml><annotation encoding="application/x-tex" id="S3.Ex4.m1.1c">s_{i+1}\;=\;T\bigl{(}s_{i},a_{i}\bigr{)},</annotation><annotation encoding="application/x-llamapun" id="S3.Ex4.m1.1d">italic_s start_POSTSUBSCRIPT italic_i + 1 end_POSTSUBSCRIPT = italic_T ( italic_s start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT , italic_a start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT ) ,</annotation></semantics></math></td> <td class="ltx_eqn_cell ltx_eqn_center_padright"></td> </tr></tbody> </table> <p class="ltx_p" id="S3.SS2.SSS2.p1.4">where <math alttext="T" class="ltx_Math" display="inline" id="S3.SS2.SSS2.p1.4.m1.1"><semantics id="S3.SS2.SSS2.p1.4.m1.1a"><mi id="S3.SS2.SSS2.p1.4.m1.1.1" xref="S3.SS2.SSS2.p1.4.m1.1.1.cmml">T</mi><annotation-xml encoding="MathML-Content" id="S3.SS2.SSS2.p1.4.m1.1b"><ci id="S3.SS2.SSS2.p1.4.m1.1.1.cmml" xref="S3.SS2.SSS2.p1.4.m1.1.1">𝑇</ci></annotation-xml><annotation encoding="application/x-tex" id="S3.SS2.SSS2.p1.4.m1.1c">T</annotation><annotation encoding="application/x-llamapun" id="S3.SS2.SSS2.p1.4.m1.1d">italic_T</annotation></semantics></math> allows:</p> <ul class="ltx_itemize" id="S3.I3"> <li class="ltx_item" id="S3.I3.i1" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item">•</span> <div class="ltx_para" id="S3.I3.i1.p1"> <p class="ltx_p" id="S3.I3.i1.p1.1">Sequential progression: <math alttext="G\to E\to S" class="ltx_Math" display="inline" id="S3.I3.i1.p1.1.m1.1"><semantics id="S3.I3.i1.p1.1.m1.1a"><mrow id="S3.I3.i1.p1.1.m1.1.1" xref="S3.I3.i1.p1.1.m1.1.1.cmml"><mi id="S3.I3.i1.p1.1.m1.1.1.2" xref="S3.I3.i1.p1.1.m1.1.1.2.cmml">G</mi><mo id="S3.I3.i1.p1.1.m1.1.1.3" stretchy="false" xref="S3.I3.i1.p1.1.m1.1.1.3.cmml">→</mo><mi id="S3.I3.i1.p1.1.m1.1.1.4" xref="S3.I3.i1.p1.1.m1.1.1.4.cmml">E</mi><mo id="S3.I3.i1.p1.1.m1.1.1.5" stretchy="false" xref="S3.I3.i1.p1.1.m1.1.1.5.cmml">→</mo><mi id="S3.I3.i1.p1.1.m1.1.1.6" xref="S3.I3.i1.p1.1.m1.1.1.6.cmml">S</mi></mrow><annotation-xml encoding="MathML-Content" id="S3.I3.i1.p1.1.m1.1b"><apply id="S3.I3.i1.p1.1.m1.1.1.cmml" xref="S3.I3.i1.p1.1.m1.1.1"><and id="S3.I3.i1.p1.1.m1.1.1a.cmml" xref="S3.I3.i1.p1.1.m1.1.1"></and><apply id="S3.I3.i1.p1.1.m1.1.1b.cmml" xref="S3.I3.i1.p1.1.m1.1.1"><ci id="S3.I3.i1.p1.1.m1.1.1.3.cmml" xref="S3.I3.i1.p1.1.m1.1.1.3">→</ci><ci id="S3.I3.i1.p1.1.m1.1.1.2.cmml" xref="S3.I3.i1.p1.1.m1.1.1.2">𝐺</ci><ci id="S3.I3.i1.p1.1.m1.1.1.4.cmml" xref="S3.I3.i1.p1.1.m1.1.1.4">𝐸</ci></apply><apply id="S3.I3.i1.p1.1.m1.1.1c.cmml" xref="S3.I3.i1.p1.1.m1.1.1"><ci id="S3.I3.i1.p1.1.m1.1.1.5.cmml" xref="S3.I3.i1.p1.1.m1.1.1.5">→</ci><share href="https://arxiv.org/html/2503.15204v1#S3.I3.i1.p1.1.m1.1.1.4.cmml" id="S3.I3.i1.p1.1.m1.1.1d.cmml" xref="S3.I3.i1.p1.1.m1.1.1"></share><ci id="S3.I3.i1.p1.1.m1.1.1.6.cmml" xref="S3.I3.i1.p1.1.m1.1.1.6">𝑆</ci></apply></apply></annotation-xml><annotation encoding="application/x-tex" id="S3.I3.i1.p1.1.m1.1c">G\to E\to S</annotation><annotation encoding="application/x-llamapun" id="S3.I3.i1.p1.1.m1.1d">italic_G → italic_E → italic_S</annotation></semantics></math>.</p> </div> </li> <li class="ltx_item" id="S3.I3.i2" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item">•</span> <div class="ltx_para" id="S3.I3.i2.p1"> <p class="ltx_p" id="S3.I3.i2.p1.1">Direct transition: <math alttext="G\to S" class="ltx_Math" display="inline" id="S3.I3.i2.p1.1.m1.1"><semantics id="S3.I3.i2.p1.1.m1.1a"><mrow id="S3.I3.i2.p1.1.m1.1.1" xref="S3.I3.i2.p1.1.m1.1.1.cmml"><mi id="S3.I3.i2.p1.1.m1.1.1.2" xref="S3.I3.i2.p1.1.m1.1.1.2.cmml">G</mi><mo id="S3.I3.i2.p1.1.m1.1.1.1" stretchy="false" xref="S3.I3.i2.p1.1.m1.1.1.1.cmml">→</mo><mi id="S3.I3.i2.p1.1.m1.1.1.3" xref="S3.I3.i2.p1.1.m1.1.1.3.cmml">S</mi></mrow><annotation-xml encoding="MathML-Content" id="S3.I3.i2.p1.1.m1.1b"><apply id="S3.I3.i2.p1.1.m1.1.1.cmml" xref="S3.I3.i2.p1.1.m1.1.1"><ci id="S3.I3.i2.p1.1.m1.1.1.1.cmml" xref="S3.I3.i2.p1.1.m1.1.1.1">→</ci><ci id="S3.I3.i2.p1.1.m1.1.1.2.cmml" xref="S3.I3.i2.p1.1.m1.1.1.2">𝐺</ci><ci id="S3.I3.i2.p1.1.m1.1.1.3.cmml" xref="S3.I3.i2.p1.1.m1.1.1.3">𝑆</ci></apply></annotation-xml><annotation encoding="application/x-tex" id="S3.I3.i2.p1.1.m1.1c">G\to S</annotation><annotation encoding="application/x-llamapun" id="S3.I3.i2.p1.1.m1.1d">italic_G → italic_S</annotation></semantics></math> if highly specific details are provided early.</p> </div> </li> <li class="ltx_item" id="S3.I3.i3" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item">•</span> <div class="ltx_para ltx_noindent" id="S3.I3.i3.p1"> <p class="ltx_p" id="S3.I3.i3.p1.1">Reversion: <math alttext="S\to G" class="ltx_Math" display="inline" id="S3.I3.i3.p1.1.m1.1"><semantics id="S3.I3.i3.p1.1.m1.1a"><mrow id="S3.I3.i3.p1.1.m1.1.1" xref="S3.I3.i3.p1.1.m1.1.1.cmml"><mi id="S3.I3.i3.p1.1.m1.1.1.2" xref="S3.I3.i3.p1.1.m1.1.1.2.cmml">S</mi><mo id="S3.I3.i3.p1.1.m1.1.1.1" stretchy="false" xref="S3.I3.i3.p1.1.m1.1.1.1.cmml">→</mo><mi id="S3.I3.i3.p1.1.m1.1.1.3" xref="S3.I3.i3.p1.1.m1.1.1.3.cmml">G</mi></mrow><annotation-xml encoding="MathML-Content" id="S3.I3.i3.p1.1.m1.1b"><apply id="S3.I3.i3.p1.1.m1.1.1.cmml" xref="S3.I3.i3.p1.1.m1.1.1"><ci id="S3.I3.i3.p1.1.m1.1.1.1.cmml" xref="S3.I3.i3.p1.1.m1.1.1.1">→</ci><ci id="S3.I3.i3.p1.1.m1.1.1.2.cmml" xref="S3.I3.i3.p1.1.m1.1.1.2">𝑆</ci><ci id="S3.I3.i3.p1.1.m1.1.1.3.cmml" xref="S3.I3.i3.p1.1.m1.1.1.3">𝐺</ci></apply></annotation-xml><annotation encoding="application/x-tex" id="S3.I3.i3.p1.1.m1.1c">S\to G</annotation><annotation encoding="application/x-llamapun" id="S3.I3.i3.p1.1.m1.1d">italic_S → italic_G</annotation></semantics></math> if crucial data is missing and turns remain.</p> </div> </li> </ul> <p class="ltx_p" id="S3.SS2.SSS2.p1.5">If the user provides no further input or the turn limit is reached, the system finalizes the collected data and initiates disease prediction (Figure <a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#S3.F1" title="Figure 1 ‣ 3 Methodology ‣ When Pigs Get Sick: Multi-Agent AI for Swine Disease Detection"><span class="ltx_text ltx_ref_tag">1</span></a>).</p> </div> </section> <section class="ltx_subsubsection" id="S3.SS2.SSS3"> <h4 class="ltx_title ltx_title_subsubsection"> <span class="ltx_tag ltx_tag_subsubsection">3.2.3 </span>Confidence-Based Disease Prediction</h4> <div class="ltx_para ltx_noindent" id="S3.SS2.SSS3.p1"> <p class="ltx_p" id="S3.SS2.SSS3.p1.3">Each specialized disease agent <math alttext="i" class="ltx_Math" display="inline" id="S3.SS2.SSS3.p1.1.m1.1"><semantics id="S3.SS2.SSS3.p1.1.m1.1a"><mi id="S3.SS2.SSS3.p1.1.m1.1.1" xref="S3.SS2.SSS3.p1.1.m1.1.1.cmml">i</mi><annotation-xml encoding="MathML-Content" id="S3.SS2.SSS3.p1.1.m1.1b"><ci id="S3.SS2.SSS3.p1.1.m1.1.1.cmml" xref="S3.SS2.SSS3.p1.1.m1.1.1">𝑖</ci></annotation-xml><annotation encoding="application/x-tex" id="S3.SS2.SSS3.p1.1.m1.1c">i</annotation><annotation encoding="application/x-llamapun" id="S3.SS2.SSS3.p1.1.m1.1d">italic_i</annotation></semantics></math> then assigns a confidence score <math alttext="p_{i}(D)" class="ltx_Math" display="inline" id="S3.SS2.SSS3.p1.2.m2.1"><semantics id="S3.SS2.SSS3.p1.2.m2.1a"><mrow id="S3.SS2.SSS3.p1.2.m2.1.2" xref="S3.SS2.SSS3.p1.2.m2.1.2.cmml"><msub id="S3.SS2.SSS3.p1.2.m2.1.2.2" xref="S3.SS2.SSS3.p1.2.m2.1.2.2.cmml"><mi id="S3.SS2.SSS3.p1.2.m2.1.2.2.2" xref="S3.SS2.SSS3.p1.2.m2.1.2.2.2.cmml">p</mi><mi id="S3.SS2.SSS3.p1.2.m2.1.2.2.3" xref="S3.SS2.SSS3.p1.2.m2.1.2.2.3.cmml">i</mi></msub><mo id="S3.SS2.SSS3.p1.2.m2.1.2.1" xref="S3.SS2.SSS3.p1.2.m2.1.2.1.cmml">⁢</mo><mrow id="S3.SS2.SSS3.p1.2.m2.1.2.3.2" xref="S3.SS2.SSS3.p1.2.m2.1.2.cmml"><mo id="S3.SS2.SSS3.p1.2.m2.1.2.3.2.1" stretchy="false" xref="S3.SS2.SSS3.p1.2.m2.1.2.cmml">(</mo><mi id="S3.SS2.SSS3.p1.2.m2.1.1" xref="S3.SS2.SSS3.p1.2.m2.1.1.cmml">D</mi><mo id="S3.SS2.SSS3.p1.2.m2.1.2.3.2.2" stretchy="false" xref="S3.SS2.SSS3.p1.2.m2.1.2.cmml">)</mo></mrow></mrow><annotation-xml encoding="MathML-Content" id="S3.SS2.SSS3.p1.2.m2.1b"><apply id="S3.SS2.SSS3.p1.2.m2.1.2.cmml" xref="S3.SS2.SSS3.p1.2.m2.1.2"><times id="S3.SS2.SSS3.p1.2.m2.1.2.1.cmml" xref="S3.SS2.SSS3.p1.2.m2.1.2.1"></times><apply id="S3.SS2.SSS3.p1.2.m2.1.2.2.cmml" xref="S3.SS2.SSS3.p1.2.m2.1.2.2"><csymbol cd="ambiguous" id="S3.SS2.SSS3.p1.2.m2.1.2.2.1.cmml" xref="S3.SS2.SSS3.p1.2.m2.1.2.2">subscript</csymbol><ci id="S3.SS2.SSS3.p1.2.m2.1.2.2.2.cmml" xref="S3.SS2.SSS3.p1.2.m2.1.2.2.2">𝑝</ci><ci id="S3.SS2.SSS3.p1.2.m2.1.2.2.3.cmml" xref="S3.SS2.SSS3.p1.2.m2.1.2.2.3">𝑖</ci></apply><ci id="S3.SS2.SSS3.p1.2.m2.1.1.cmml" xref="S3.SS2.SSS3.p1.2.m2.1.1">𝐷</ci></apply></annotation-xml><annotation encoding="application/x-tex" id="S3.SS2.SSS3.p1.2.m2.1c">p_{i}(D)</annotation><annotation encoding="application/x-llamapun" id="S3.SS2.SSS3.p1.2.m2.1d">italic_p start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT ( italic_D )</annotation></semantics></math> for disease <math alttext="D" class="ltx_Math" display="inline" id="S3.SS2.SSS3.p1.3.m3.1"><semantics id="S3.SS2.SSS3.p1.3.m3.1a"><mi id="S3.SS2.SSS3.p1.3.m3.1.1" xref="S3.SS2.SSS3.p1.3.m3.1.1.cmml">D</mi><annotation-xml encoding="MathML-Content" id="S3.SS2.SSS3.p1.3.m3.1b"><ci id="S3.SS2.SSS3.p1.3.m3.1.1.cmml" xref="S3.SS2.SSS3.p1.3.m3.1.1">𝐷</ci></annotation-xml><annotation encoding="application/x-tex" id="S3.SS2.SSS3.p1.3.m3.1c">D</annotation><annotation encoding="application/x-llamapun" id="S3.SS2.SSS3.p1.3.m3.1d">italic_D</annotation></semantics></math>, aggregated as:</p> </div> <div class="ltx_para ltx_noindent" id="S3.SS2.SSS3.p2"> <table class="ltx_equation ltx_eqn_table" id="S3.Ex5"> <tbody><tr class="ltx_equation ltx_eqn_row ltx_align_baseline"> <td class="ltx_eqn_cell ltx_eqn_center_padleft"></td> <td class="ltx_eqn_cell ltx_align_center"><math alttext="C(D)\;=\;\sum_{i=1}^{n}\alpha_{i}\,p_{i}(D)," class="ltx_Math" display="block" id="S3.Ex5.m1.3"><semantics id="S3.Ex5.m1.3a"><mrow id="S3.Ex5.m1.3.3.1" xref="S3.Ex5.m1.3.3.1.1.cmml"><mrow id="S3.Ex5.m1.3.3.1.1" xref="S3.Ex5.m1.3.3.1.1.cmml"><mrow id="S3.Ex5.m1.3.3.1.1.2" xref="S3.Ex5.m1.3.3.1.1.2.cmml"><mi id="S3.Ex5.m1.3.3.1.1.2.2" xref="S3.Ex5.m1.3.3.1.1.2.2.cmml">C</mi><mo id="S3.Ex5.m1.3.3.1.1.2.1" xref="S3.Ex5.m1.3.3.1.1.2.1.cmml">⁢</mo><mrow id="S3.Ex5.m1.3.3.1.1.2.3.2" xref="S3.Ex5.m1.3.3.1.1.2.cmml"><mo id="S3.Ex5.m1.3.3.1.1.2.3.2.1" stretchy="false" xref="S3.Ex5.m1.3.3.1.1.2.cmml">(</mo><mi id="S3.Ex5.m1.1.1" xref="S3.Ex5.m1.1.1.cmml">D</mi><mo id="S3.Ex5.m1.3.3.1.1.2.3.2.2" rspace="0.280em" stretchy="false" xref="S3.Ex5.m1.3.3.1.1.2.cmml">)</mo></mrow></mrow><mo id="S3.Ex5.m1.3.3.1.1.1" rspace="0.391em" xref="S3.Ex5.m1.3.3.1.1.1.cmml">=</mo><mrow id="S3.Ex5.m1.3.3.1.1.3" xref="S3.Ex5.m1.3.3.1.1.3.cmml"><munderover id="S3.Ex5.m1.3.3.1.1.3.1" xref="S3.Ex5.m1.3.3.1.1.3.1.cmml"><mo id="S3.Ex5.m1.3.3.1.1.3.1.2.2" movablelimits="false" xref="S3.Ex5.m1.3.3.1.1.3.1.2.2.cmml">∑</mo><mrow id="S3.Ex5.m1.3.3.1.1.3.1.2.3" xref="S3.Ex5.m1.3.3.1.1.3.1.2.3.cmml"><mi id="S3.Ex5.m1.3.3.1.1.3.1.2.3.2" xref="S3.Ex5.m1.3.3.1.1.3.1.2.3.2.cmml">i</mi><mo id="S3.Ex5.m1.3.3.1.1.3.1.2.3.1" xref="S3.Ex5.m1.3.3.1.1.3.1.2.3.1.cmml">=</mo><mn id="S3.Ex5.m1.3.3.1.1.3.1.2.3.3" xref="S3.Ex5.m1.3.3.1.1.3.1.2.3.3.cmml">1</mn></mrow><mi id="S3.Ex5.m1.3.3.1.1.3.1.3" xref="S3.Ex5.m1.3.3.1.1.3.1.3.cmml">n</mi></munderover><mrow id="S3.Ex5.m1.3.3.1.1.3.2" xref="S3.Ex5.m1.3.3.1.1.3.2.cmml"><msub id="S3.Ex5.m1.3.3.1.1.3.2.2" xref="S3.Ex5.m1.3.3.1.1.3.2.2.cmml"><mi id="S3.Ex5.m1.3.3.1.1.3.2.2.2" xref="S3.Ex5.m1.3.3.1.1.3.2.2.2.cmml">α</mi><mi id="S3.Ex5.m1.3.3.1.1.3.2.2.3" xref="S3.Ex5.m1.3.3.1.1.3.2.2.3.cmml">i</mi></msub><mo id="S3.Ex5.m1.3.3.1.1.3.2.1" xref="S3.Ex5.m1.3.3.1.1.3.2.1.cmml">⁢</mo><msub id="S3.Ex5.m1.3.3.1.1.3.2.3" xref="S3.Ex5.m1.3.3.1.1.3.2.3.cmml"><mi id="S3.Ex5.m1.3.3.1.1.3.2.3.2" xref="S3.Ex5.m1.3.3.1.1.3.2.3.2.cmml">p</mi><mi id="S3.Ex5.m1.3.3.1.1.3.2.3.3" xref="S3.Ex5.m1.3.3.1.1.3.2.3.3.cmml">i</mi></msub><mo id="S3.Ex5.m1.3.3.1.1.3.2.1a" xref="S3.Ex5.m1.3.3.1.1.3.2.1.cmml">⁢</mo><mrow id="S3.Ex5.m1.3.3.1.1.3.2.4.2" xref="S3.Ex5.m1.3.3.1.1.3.2.cmml"><mo id="S3.Ex5.m1.3.3.1.1.3.2.4.2.1" stretchy="false" xref="S3.Ex5.m1.3.3.1.1.3.2.cmml">(</mo><mi id="S3.Ex5.m1.2.2" xref="S3.Ex5.m1.2.2.cmml">D</mi><mo id="S3.Ex5.m1.3.3.1.1.3.2.4.2.2" stretchy="false" xref="S3.Ex5.m1.3.3.1.1.3.2.cmml">)</mo></mrow></mrow></mrow></mrow><mo id="S3.Ex5.m1.3.3.1.2" xref="S3.Ex5.m1.3.3.1.1.cmml">,</mo></mrow><annotation-xml encoding="MathML-Content" id="S3.Ex5.m1.3b"><apply id="S3.Ex5.m1.3.3.1.1.cmml" xref="S3.Ex5.m1.3.3.1"><eq id="S3.Ex5.m1.3.3.1.1.1.cmml" xref="S3.Ex5.m1.3.3.1.1.1"></eq><apply id="S3.Ex5.m1.3.3.1.1.2.cmml" xref="S3.Ex5.m1.3.3.1.1.2"><times id="S3.Ex5.m1.3.3.1.1.2.1.cmml" xref="S3.Ex5.m1.3.3.1.1.2.1"></times><ci id="S3.Ex5.m1.3.3.1.1.2.2.cmml" xref="S3.Ex5.m1.3.3.1.1.2.2">𝐶</ci><ci id="S3.Ex5.m1.1.1.cmml" xref="S3.Ex5.m1.1.1">𝐷</ci></apply><apply id="S3.Ex5.m1.3.3.1.1.3.cmml" xref="S3.Ex5.m1.3.3.1.1.3"><apply id="S3.Ex5.m1.3.3.1.1.3.1.cmml" xref="S3.Ex5.m1.3.3.1.1.3.1"><csymbol cd="ambiguous" id="S3.Ex5.m1.3.3.1.1.3.1.1.cmml" xref="S3.Ex5.m1.3.3.1.1.3.1">superscript</csymbol><apply id="S3.Ex5.m1.3.3.1.1.3.1.2.cmml" xref="S3.Ex5.m1.3.3.1.1.3.1"><csymbol cd="ambiguous" id="S3.Ex5.m1.3.3.1.1.3.1.2.1.cmml" xref="S3.Ex5.m1.3.3.1.1.3.1">subscript</csymbol><sum id="S3.Ex5.m1.3.3.1.1.3.1.2.2.cmml" xref="S3.Ex5.m1.3.3.1.1.3.1.2.2"></sum><apply id="S3.Ex5.m1.3.3.1.1.3.1.2.3.cmml" xref="S3.Ex5.m1.3.3.1.1.3.1.2.3"><eq id="S3.Ex5.m1.3.3.1.1.3.1.2.3.1.cmml" xref="S3.Ex5.m1.3.3.1.1.3.1.2.3.1"></eq><ci id="S3.Ex5.m1.3.3.1.1.3.1.2.3.2.cmml" xref="S3.Ex5.m1.3.3.1.1.3.1.2.3.2">𝑖</ci><cn id="S3.Ex5.m1.3.3.1.1.3.1.2.3.3.cmml" type="integer" xref="S3.Ex5.m1.3.3.1.1.3.1.2.3.3">1</cn></apply></apply><ci id="S3.Ex5.m1.3.3.1.1.3.1.3.cmml" xref="S3.Ex5.m1.3.3.1.1.3.1.3">𝑛</ci></apply><apply id="S3.Ex5.m1.3.3.1.1.3.2.cmml" xref="S3.Ex5.m1.3.3.1.1.3.2"><times id="S3.Ex5.m1.3.3.1.1.3.2.1.cmml" xref="S3.Ex5.m1.3.3.1.1.3.2.1"></times><apply id="S3.Ex5.m1.3.3.1.1.3.2.2.cmml" xref="S3.Ex5.m1.3.3.1.1.3.2.2"><csymbol cd="ambiguous" id="S3.Ex5.m1.3.3.1.1.3.2.2.1.cmml" xref="S3.Ex5.m1.3.3.1.1.3.2.2">subscript</csymbol><ci id="S3.Ex5.m1.3.3.1.1.3.2.2.2.cmml" xref="S3.Ex5.m1.3.3.1.1.3.2.2.2">𝛼</ci><ci id="S3.Ex5.m1.3.3.1.1.3.2.2.3.cmml" xref="S3.Ex5.m1.3.3.1.1.3.2.2.3">𝑖</ci></apply><apply id="S3.Ex5.m1.3.3.1.1.3.2.3.cmml" xref="S3.Ex5.m1.3.3.1.1.3.2.3"><csymbol cd="ambiguous" id="S3.Ex5.m1.3.3.1.1.3.2.3.1.cmml" xref="S3.Ex5.m1.3.3.1.1.3.2.3">subscript</csymbol><ci id="S3.Ex5.m1.3.3.1.1.3.2.3.2.cmml" xref="S3.Ex5.m1.3.3.1.1.3.2.3.2">𝑝</ci><ci id="S3.Ex5.m1.3.3.1.1.3.2.3.3.cmml" xref="S3.Ex5.m1.3.3.1.1.3.2.3.3">𝑖</ci></apply><ci id="S3.Ex5.m1.2.2.cmml" xref="S3.Ex5.m1.2.2">𝐷</ci></apply></apply></apply></annotation-xml><annotation encoding="application/x-tex" id="S3.Ex5.m1.3c">C(D)\;=\;\sum_{i=1}^{n}\alpha_{i}\,p_{i}(D),</annotation><annotation encoding="application/x-llamapun" id="S3.Ex5.m1.3d">italic_C ( italic_D ) = ∑ start_POSTSUBSCRIPT italic_i = 1 end_POSTSUBSCRIPT start_POSTSUPERSCRIPT italic_n end_POSTSUPERSCRIPT italic_α start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT italic_p start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT ( italic_D ) ,</annotation></semantics></math></td> <td class="ltx_eqn_cell ltx_eqn_center_padright"></td> </tr></tbody> </table> </div> <div class="ltx_para ltx_noindent" id="S3.SS2.SSS3.p3"> <p class="ltx_p" id="S3.SS2.SSS3.p3.3">where <math alttext="\alpha_{i}" class="ltx_Math" display="inline" id="S3.SS2.SSS3.p3.1.m1.1"><semantics id="S3.SS2.SSS3.p3.1.m1.1a"><msub id="S3.SS2.SSS3.p3.1.m1.1.1" xref="S3.SS2.SSS3.p3.1.m1.1.1.cmml"><mi id="S3.SS2.SSS3.p3.1.m1.1.1.2" xref="S3.SS2.SSS3.p3.1.m1.1.1.2.cmml">α</mi><mi id="S3.SS2.SSS3.p3.1.m1.1.1.3" xref="S3.SS2.SSS3.p3.1.m1.1.1.3.cmml">i</mi></msub><annotation-xml encoding="MathML-Content" id="S3.SS2.SSS3.p3.1.m1.1b"><apply id="S3.SS2.SSS3.p3.1.m1.1.1.cmml" xref="S3.SS2.SSS3.p3.1.m1.1.1"><csymbol cd="ambiguous" id="S3.SS2.SSS3.p3.1.m1.1.1.1.cmml" xref="S3.SS2.SSS3.p3.1.m1.1.1">subscript</csymbol><ci id="S3.SS2.SSS3.p3.1.m1.1.1.2.cmml" xref="S3.SS2.SSS3.p3.1.m1.1.1.2">𝛼</ci><ci id="S3.SS2.SSS3.p3.1.m1.1.1.3.cmml" xref="S3.SS2.SSS3.p3.1.m1.1.1.3">𝑖</ci></apply></annotation-xml><annotation encoding="application/x-tex" id="S3.SS2.SSS3.p3.1.m1.1c">\alpha_{i}</annotation><annotation encoding="application/x-llamapun" id="S3.SS2.SSS3.p3.1.m1.1d">italic_α start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT</annotation></semantics></math> represents the weight assigned to agent <math alttext="i" class="ltx_Math" display="inline" id="S3.SS2.SSS3.p3.2.m2.1"><semantics id="S3.SS2.SSS3.p3.2.m2.1a"><mi id="S3.SS2.SSS3.p3.2.m2.1.1" xref="S3.SS2.SSS3.p3.2.m2.1.1.cmml">i</mi><annotation-xml encoding="MathML-Content" id="S3.SS2.SSS3.p3.2.m2.1b"><ci id="S3.SS2.SSS3.p3.2.m2.1.1.cmml" xref="S3.SS2.SSS3.p3.2.m2.1.1">𝑖</ci></annotation-xml><annotation encoding="application/x-tex" id="S3.SS2.SSS3.p3.2.m2.1c">i</annotation><annotation encoding="application/x-llamapun" id="S3.SS2.SSS3.p3.2.m2.1d">italic_i</annotation></semantics></math>, typically reflecting its expertise or reliability. A disease is predicted when its confidence score meets or exceeds the predefined threshold <math alttext="\tau" class="ltx_Math" display="inline" id="S3.SS2.SSS3.p3.3.m3.1"><semantics id="S3.SS2.SSS3.p3.3.m3.1a"><mi id="S3.SS2.SSS3.p3.3.m3.1.1" xref="S3.SS2.SSS3.p3.3.m3.1.1.cmml">τ</mi><annotation-xml encoding="MathML-Content" id="S3.SS2.SSS3.p3.3.m3.1b"><ci id="S3.SS2.SSS3.p3.3.m3.1.1.cmml" xref="S3.SS2.SSS3.p3.3.m3.1.1">𝜏</ci></annotation-xml><annotation encoding="application/x-tex" id="S3.SS2.SSS3.p3.3.m3.1c">\tau</annotation><annotation encoding="application/x-llamapun" id="S3.SS2.SSS3.p3.3.m3.1d">italic_τ</annotation></semantics></math>:</p> </div> <div class="ltx_para ltx_noindent" id="S3.SS2.SSS3.p4"> <table class="ltx_equation ltx_eqn_table" id="S3.Ex6"> <tbody><tr class="ltx_equation ltx_eqn_row ltx_align_baseline"> <td class="ltx_eqn_cell ltx_eqn_center_padleft"></td> <td class="ltx_eqn_cell ltx_align_center"><math alttext="\widehat{D}\;=\;\{\,D\mid C(D)\,\geq\,\tau\}." class="ltx_Math" display="block" id="S3.Ex6.m1.3"><semantics id="S3.Ex6.m1.3a"><mrow id="S3.Ex6.m1.3.3.1" xref="S3.Ex6.m1.3.3.1.1.cmml"><mrow id="S3.Ex6.m1.3.3.1.1" xref="S3.Ex6.m1.3.3.1.1.cmml"><mover accent="true" id="S3.Ex6.m1.3.3.1.1.3" xref="S3.Ex6.m1.3.3.1.1.3.cmml"><mi id="S3.Ex6.m1.3.3.1.1.3.2" xref="S3.Ex6.m1.3.3.1.1.3.2.cmml">D</mi><mo id="S3.Ex6.m1.3.3.1.1.3.1" xref="S3.Ex6.m1.3.3.1.1.3.1.cmml">^</mo></mover><mo id="S3.Ex6.m1.3.3.1.1.2" lspace="0.558em" rspace="0.558em" xref="S3.Ex6.m1.3.3.1.1.2.cmml">=</mo><mrow id="S3.Ex6.m1.3.3.1.1.1.1" xref="S3.Ex6.m1.3.3.1.1.1.2.cmml"><mo id="S3.Ex6.m1.3.3.1.1.1.1.2" rspace="0.170em" stretchy="false" xref="S3.Ex6.m1.3.3.1.1.1.2.1.cmml">{</mo><mi id="S3.Ex6.m1.2.2" xref="S3.Ex6.m1.2.2.cmml">D</mi><mo fence="true" id="S3.Ex6.m1.3.3.1.1.1.1.3" lspace="0em" rspace="0em" xref="S3.Ex6.m1.3.3.1.1.1.2.1.cmml">∣</mo><mrow id="S3.Ex6.m1.3.3.1.1.1.1.1" xref="S3.Ex6.m1.3.3.1.1.1.1.1.cmml"><mrow id="S3.Ex6.m1.3.3.1.1.1.1.1.2" xref="S3.Ex6.m1.3.3.1.1.1.1.1.2.cmml"><mi id="S3.Ex6.m1.3.3.1.1.1.1.1.2.2" xref="S3.Ex6.m1.3.3.1.1.1.1.1.2.2.cmml">C</mi><mo id="S3.Ex6.m1.3.3.1.1.1.1.1.2.1" xref="S3.Ex6.m1.3.3.1.1.1.1.1.2.1.cmml">⁢</mo><mrow id="S3.Ex6.m1.3.3.1.1.1.1.1.2.3.2" xref="S3.Ex6.m1.3.3.1.1.1.1.1.2.cmml"><mo id="S3.Ex6.m1.3.3.1.1.1.1.1.2.3.2.1" stretchy="false" xref="S3.Ex6.m1.3.3.1.1.1.1.1.2.cmml">(</mo><mi id="S3.Ex6.m1.1.1" xref="S3.Ex6.m1.1.1.cmml">D</mi><mo id="S3.Ex6.m1.3.3.1.1.1.1.1.2.3.2.2" rspace="0.170em" stretchy="false" xref="S3.Ex6.m1.3.3.1.1.1.1.1.2.cmml">)</mo></mrow></mrow><mo id="S3.Ex6.m1.3.3.1.1.1.1.1.1" rspace="0.448em" xref="S3.Ex6.m1.3.3.1.1.1.1.1.1.cmml">≥</mo><mi id="S3.Ex6.m1.3.3.1.1.1.1.1.3" xref="S3.Ex6.m1.3.3.1.1.1.1.1.3.cmml">τ</mi></mrow><mo id="S3.Ex6.m1.3.3.1.1.1.1.4" stretchy="false" xref="S3.Ex6.m1.3.3.1.1.1.2.1.cmml">}</mo></mrow></mrow><mo id="S3.Ex6.m1.3.3.1.2" lspace="0em" xref="S3.Ex6.m1.3.3.1.1.cmml">.</mo></mrow><annotation-xml encoding="MathML-Content" id="S3.Ex6.m1.3b"><apply id="S3.Ex6.m1.3.3.1.1.cmml" xref="S3.Ex6.m1.3.3.1"><eq id="S3.Ex6.m1.3.3.1.1.2.cmml" xref="S3.Ex6.m1.3.3.1.1.2"></eq><apply id="S3.Ex6.m1.3.3.1.1.3.cmml" xref="S3.Ex6.m1.3.3.1.1.3"><ci id="S3.Ex6.m1.3.3.1.1.3.1.cmml" xref="S3.Ex6.m1.3.3.1.1.3.1">^</ci><ci id="S3.Ex6.m1.3.3.1.1.3.2.cmml" xref="S3.Ex6.m1.3.3.1.1.3.2">𝐷</ci></apply><apply id="S3.Ex6.m1.3.3.1.1.1.2.cmml" xref="S3.Ex6.m1.3.3.1.1.1.1"><csymbol cd="latexml" id="S3.Ex6.m1.3.3.1.1.1.2.1.cmml" xref="S3.Ex6.m1.3.3.1.1.1.1.2">conditional-set</csymbol><ci id="S3.Ex6.m1.2.2.cmml" xref="S3.Ex6.m1.2.2">𝐷</ci><apply id="S3.Ex6.m1.3.3.1.1.1.1.1.cmml" xref="S3.Ex6.m1.3.3.1.1.1.1.1"><geq id="S3.Ex6.m1.3.3.1.1.1.1.1.1.cmml" xref="S3.Ex6.m1.3.3.1.1.1.1.1.1"></geq><apply id="S3.Ex6.m1.3.3.1.1.1.1.1.2.cmml" xref="S3.Ex6.m1.3.3.1.1.1.1.1.2"><times id="S3.Ex6.m1.3.3.1.1.1.1.1.2.1.cmml" xref="S3.Ex6.m1.3.3.1.1.1.1.1.2.1"></times><ci id="S3.Ex6.m1.3.3.1.1.1.1.1.2.2.cmml" xref="S3.Ex6.m1.3.3.1.1.1.1.1.2.2">𝐶</ci><ci id="S3.Ex6.m1.1.1.cmml" xref="S3.Ex6.m1.1.1">𝐷</ci></apply><ci id="S3.Ex6.m1.3.3.1.1.1.1.1.3.cmml" xref="S3.Ex6.m1.3.3.1.1.1.1.1.3">𝜏</ci></apply></apply></apply></annotation-xml><annotation encoding="application/x-tex" id="S3.Ex6.m1.3c">\widehat{D}\;=\;\{\,D\mid C(D)\,\geq\,\tau\}.</annotation><annotation encoding="application/x-llamapun" id="S3.Ex6.m1.3d">over^ start_ARG italic_D end_ARG = { italic_D ∣ italic_C ( italic_D ) ≥ italic_τ } .</annotation></semantics></math></td> <td class="ltx_eqn_cell ltx_eqn_center_padright"></td> </tr></tbody> </table> </div> <div class="ltx_para ltx_noindent" id="S3.SS2.SSS3.p5"> <p class="ltx_p" id="S3.SS2.SSS3.p5.1">The system categorizes predictions into four confidence tiers to determine the appropriate response:</p> <ul class="ltx_itemize" id="S3.I4"> <li class="ltx_item" id="S3.I4.i1" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item">•</span> <div class="ltx_para" id="S3.I4.i1.p1"> <p class="ltx_p" id="S3.I4.i1.p1.1"><span class="ltx_text ltx_font_bold" id="S3.I4.i1.p1.1.1">Very High Confidence</span> (<math alttext="C(D)\geq 0.75" class="ltx_Math" display="inline" id="S3.I4.i1.p1.1.m1.1"><semantics id="S3.I4.i1.p1.1.m1.1a"><mrow id="S3.I4.i1.p1.1.m1.1.2" xref="S3.I4.i1.p1.1.m1.1.2.cmml"><mrow id="S3.I4.i1.p1.1.m1.1.2.2" xref="S3.I4.i1.p1.1.m1.1.2.2.cmml"><mi id="S3.I4.i1.p1.1.m1.1.2.2.2" xref="S3.I4.i1.p1.1.m1.1.2.2.2.cmml">C</mi><mo id="S3.I4.i1.p1.1.m1.1.2.2.1" xref="S3.I4.i1.p1.1.m1.1.2.2.1.cmml">⁢</mo><mrow id="S3.I4.i1.p1.1.m1.1.2.2.3.2" xref="S3.I4.i1.p1.1.m1.1.2.2.cmml"><mo id="S3.I4.i1.p1.1.m1.1.2.2.3.2.1" stretchy="false" xref="S3.I4.i1.p1.1.m1.1.2.2.cmml">(</mo><mi id="S3.I4.i1.p1.1.m1.1.1" xref="S3.I4.i1.p1.1.m1.1.1.cmml">D</mi><mo id="S3.I4.i1.p1.1.m1.1.2.2.3.2.2" stretchy="false" xref="S3.I4.i1.p1.1.m1.1.2.2.cmml">)</mo></mrow></mrow><mo id="S3.I4.i1.p1.1.m1.1.2.1" xref="S3.I4.i1.p1.1.m1.1.2.1.cmml">≥</mo><mn id="S3.I4.i1.p1.1.m1.1.2.3" xref="S3.I4.i1.p1.1.m1.1.2.3.cmml">0.75</mn></mrow><annotation-xml encoding="MathML-Content" id="S3.I4.i1.p1.1.m1.1b"><apply id="S3.I4.i1.p1.1.m1.1.2.cmml" xref="S3.I4.i1.p1.1.m1.1.2"><geq id="S3.I4.i1.p1.1.m1.1.2.1.cmml" xref="S3.I4.i1.p1.1.m1.1.2.1"></geq><apply id="S3.I4.i1.p1.1.m1.1.2.2.cmml" xref="S3.I4.i1.p1.1.m1.1.2.2"><times id="S3.I4.i1.p1.1.m1.1.2.2.1.cmml" xref="S3.I4.i1.p1.1.m1.1.2.2.1"></times><ci id="S3.I4.i1.p1.1.m1.1.2.2.2.cmml" xref="S3.I4.i1.p1.1.m1.1.2.2.2">𝐶</ci><ci id="S3.I4.i1.p1.1.m1.1.1.cmml" xref="S3.I4.i1.p1.1.m1.1.1">𝐷</ci></apply><cn id="S3.I4.i1.p1.1.m1.1.2.3.cmml" type="float" xref="S3.I4.i1.p1.1.m1.1.2.3">0.75</cn></apply></annotation-xml><annotation encoding="application/x-tex" id="S3.I4.i1.p1.1.m1.1c">C(D)\geq 0.75</annotation><annotation encoding="application/x-llamapun" id="S3.I4.i1.p1.1.m1.1d">italic_C ( italic_D ) ≥ 0.75</annotation></semantics></math>): Strong predictive certainty; prioritized for action.</p> </div> </li> <li class="ltx_item" id="S3.I4.i2" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item">•</span> <div class="ltx_para" id="S3.I4.i2.p1"> <p class="ltx_p" id="S3.I4.i2.p1.1"><span class="ltx_text ltx_font_bold" id="S3.I4.i2.p1.1.1">High Confidence</span> (<math alttext="0.624\leq C(D)<0.75" class="ltx_Math" display="inline" id="S3.I4.i2.p1.1.m1.1"><semantics id="S3.I4.i2.p1.1.m1.1a"><mrow id="S3.I4.i2.p1.1.m1.1.2" xref="S3.I4.i2.p1.1.m1.1.2.cmml"><mn id="S3.I4.i2.p1.1.m1.1.2.2" xref="S3.I4.i2.p1.1.m1.1.2.2.cmml">0.624</mn><mo id="S3.I4.i2.p1.1.m1.1.2.3" xref="S3.I4.i2.p1.1.m1.1.2.3.cmml">≤</mo><mrow id="S3.I4.i2.p1.1.m1.1.2.4" xref="S3.I4.i2.p1.1.m1.1.2.4.cmml"><mi id="S3.I4.i2.p1.1.m1.1.2.4.2" xref="S3.I4.i2.p1.1.m1.1.2.4.2.cmml">C</mi><mo id="S3.I4.i2.p1.1.m1.1.2.4.1" xref="S3.I4.i2.p1.1.m1.1.2.4.1.cmml">⁢</mo><mrow id="S3.I4.i2.p1.1.m1.1.2.4.3.2" xref="S3.I4.i2.p1.1.m1.1.2.4.cmml"><mo id="S3.I4.i2.p1.1.m1.1.2.4.3.2.1" stretchy="false" xref="S3.I4.i2.p1.1.m1.1.2.4.cmml">(</mo><mi id="S3.I4.i2.p1.1.m1.1.1" xref="S3.I4.i2.p1.1.m1.1.1.cmml">D</mi><mo id="S3.I4.i2.p1.1.m1.1.2.4.3.2.2" stretchy="false" xref="S3.I4.i2.p1.1.m1.1.2.4.cmml">)</mo></mrow></mrow><mo id="S3.I4.i2.p1.1.m1.1.2.5" xref="S3.I4.i2.p1.1.m1.1.2.5.cmml"><</mo><mn id="S3.I4.i2.p1.1.m1.1.2.6" xref="S3.I4.i2.p1.1.m1.1.2.6.cmml">0.75</mn></mrow><annotation-xml encoding="MathML-Content" id="S3.I4.i2.p1.1.m1.1b"><apply id="S3.I4.i2.p1.1.m1.1.2.cmml" xref="S3.I4.i2.p1.1.m1.1.2"><and id="S3.I4.i2.p1.1.m1.1.2a.cmml" xref="S3.I4.i2.p1.1.m1.1.2"></and><apply id="S3.I4.i2.p1.1.m1.1.2b.cmml" xref="S3.I4.i2.p1.1.m1.1.2"><leq id="S3.I4.i2.p1.1.m1.1.2.3.cmml" xref="S3.I4.i2.p1.1.m1.1.2.3"></leq><cn id="S3.I4.i2.p1.1.m1.1.2.2.cmml" type="float" xref="S3.I4.i2.p1.1.m1.1.2.2">0.624</cn><apply id="S3.I4.i2.p1.1.m1.1.2.4.cmml" xref="S3.I4.i2.p1.1.m1.1.2.4"><times id="S3.I4.i2.p1.1.m1.1.2.4.1.cmml" xref="S3.I4.i2.p1.1.m1.1.2.4.1"></times><ci id="S3.I4.i2.p1.1.m1.1.2.4.2.cmml" xref="S3.I4.i2.p1.1.m1.1.2.4.2">𝐶</ci><ci id="S3.I4.i2.p1.1.m1.1.1.cmml" xref="S3.I4.i2.p1.1.m1.1.1">𝐷</ci></apply></apply><apply id="S3.I4.i2.p1.1.m1.1.2c.cmml" xref="S3.I4.i2.p1.1.m1.1.2"><lt id="S3.I4.i2.p1.1.m1.1.2.5.cmml" xref="S3.I4.i2.p1.1.m1.1.2.5"></lt><share href="https://arxiv.org/html/2503.15204v1#S3.I4.i2.p1.1.m1.1.2.4.cmml" id="S3.I4.i2.p1.1.m1.1.2d.cmml" xref="S3.I4.i2.p1.1.m1.1.2"></share><cn id="S3.I4.i2.p1.1.m1.1.2.6.cmml" type="float" xref="S3.I4.i2.p1.1.m1.1.2.6">0.75</cn></apply></apply></annotation-xml><annotation encoding="application/x-tex" id="S3.I4.i2.p1.1.m1.1c">0.624\leq C(D)<0.75</annotation><annotation encoding="application/x-llamapun" id="S3.I4.i2.p1.1.m1.1d">0.624 ≤ italic_C ( italic_D ) < 0.75</annotation></semantics></math>): Highly reliable but may require additional validation.</p> </div> </li> <li class="ltx_item" id="S3.I4.i3" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item">•</span> <div class="ltx_para" id="S3.I4.i3.p1"> <p class="ltx_p" id="S3.I4.i3.p1.1"><span class="ltx_text ltx_font_bold" id="S3.I4.i3.p1.1.1">Medium Confidence</span> (<math alttext="0.375\leq C(D)<0.624" class="ltx_Math" display="inline" id="S3.I4.i3.p1.1.m1.1"><semantics id="S3.I4.i3.p1.1.m1.1a"><mrow id="S3.I4.i3.p1.1.m1.1.2" xref="S3.I4.i3.p1.1.m1.1.2.cmml"><mn id="S3.I4.i3.p1.1.m1.1.2.2" xref="S3.I4.i3.p1.1.m1.1.2.2.cmml">0.375</mn><mo id="S3.I4.i3.p1.1.m1.1.2.3" xref="S3.I4.i3.p1.1.m1.1.2.3.cmml">≤</mo><mrow id="S3.I4.i3.p1.1.m1.1.2.4" xref="S3.I4.i3.p1.1.m1.1.2.4.cmml"><mi id="S3.I4.i3.p1.1.m1.1.2.4.2" xref="S3.I4.i3.p1.1.m1.1.2.4.2.cmml">C</mi><mo id="S3.I4.i3.p1.1.m1.1.2.4.1" xref="S3.I4.i3.p1.1.m1.1.2.4.1.cmml">⁢</mo><mrow id="S3.I4.i3.p1.1.m1.1.2.4.3.2" xref="S3.I4.i3.p1.1.m1.1.2.4.cmml"><mo id="S3.I4.i3.p1.1.m1.1.2.4.3.2.1" stretchy="false" xref="S3.I4.i3.p1.1.m1.1.2.4.cmml">(</mo><mi id="S3.I4.i3.p1.1.m1.1.1" xref="S3.I4.i3.p1.1.m1.1.1.cmml">D</mi><mo id="S3.I4.i3.p1.1.m1.1.2.4.3.2.2" stretchy="false" xref="S3.I4.i3.p1.1.m1.1.2.4.cmml">)</mo></mrow></mrow><mo id="S3.I4.i3.p1.1.m1.1.2.5" xref="S3.I4.i3.p1.1.m1.1.2.5.cmml"><</mo><mn id="S3.I4.i3.p1.1.m1.1.2.6" xref="S3.I4.i3.p1.1.m1.1.2.6.cmml">0.624</mn></mrow><annotation-xml encoding="MathML-Content" id="S3.I4.i3.p1.1.m1.1b"><apply id="S3.I4.i3.p1.1.m1.1.2.cmml" xref="S3.I4.i3.p1.1.m1.1.2"><and id="S3.I4.i3.p1.1.m1.1.2a.cmml" xref="S3.I4.i3.p1.1.m1.1.2"></and><apply id="S3.I4.i3.p1.1.m1.1.2b.cmml" xref="S3.I4.i3.p1.1.m1.1.2"><leq id="S3.I4.i3.p1.1.m1.1.2.3.cmml" xref="S3.I4.i3.p1.1.m1.1.2.3"></leq><cn id="S3.I4.i3.p1.1.m1.1.2.2.cmml" type="float" xref="S3.I4.i3.p1.1.m1.1.2.2">0.375</cn><apply id="S3.I4.i3.p1.1.m1.1.2.4.cmml" xref="S3.I4.i3.p1.1.m1.1.2.4"><times id="S3.I4.i3.p1.1.m1.1.2.4.1.cmml" xref="S3.I4.i3.p1.1.m1.1.2.4.1"></times><ci id="S3.I4.i3.p1.1.m1.1.2.4.2.cmml" xref="S3.I4.i3.p1.1.m1.1.2.4.2">𝐶</ci><ci id="S3.I4.i3.p1.1.m1.1.1.cmml" xref="S3.I4.i3.p1.1.m1.1.1">𝐷</ci></apply></apply><apply id="S3.I4.i3.p1.1.m1.1.2c.cmml" xref="S3.I4.i3.p1.1.m1.1.2"><lt id="S3.I4.i3.p1.1.m1.1.2.5.cmml" xref="S3.I4.i3.p1.1.m1.1.2.5"></lt><share href="https://arxiv.org/html/2503.15204v1#S3.I4.i3.p1.1.m1.1.2.4.cmml" id="S3.I4.i3.p1.1.m1.1.2d.cmml" xref="S3.I4.i3.p1.1.m1.1.2"></share><cn id="S3.I4.i3.p1.1.m1.1.2.6.cmml" type="float" xref="S3.I4.i3.p1.1.m1.1.2.6">0.624</cn></apply></apply></annotation-xml><annotation encoding="application/x-tex" id="S3.I4.i3.p1.1.m1.1c">0.375\leq C(D)<0.624</annotation><annotation encoding="application/x-llamapun" id="S3.I4.i3.p1.1.m1.1d">0.375 ≤ italic_C ( italic_D ) < 0.624</annotation></semantics></math>): Moderate likelihood; further assessment is advised.</p> </div> </li> <li class="ltx_item" id="S3.I4.i4" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item">•</span> <div class="ltx_para ltx_noindent" id="S3.I4.i4.p1"> <p class="ltx_p" id="S3.I4.i4.p1.1"><span class="ltx_text ltx_font_bold" id="S3.I4.i4.p1.1.1">Low Confidence</span> (<math alttext="C(D)<0.375" class="ltx_Math" display="inline" id="S3.I4.i4.p1.1.m1.1"><semantics id="S3.I4.i4.p1.1.m1.1a"><mrow id="S3.I4.i4.p1.1.m1.1.2" xref="S3.I4.i4.p1.1.m1.1.2.cmml"><mrow id="S3.I4.i4.p1.1.m1.1.2.2" xref="S3.I4.i4.p1.1.m1.1.2.2.cmml"><mi id="S3.I4.i4.p1.1.m1.1.2.2.2" xref="S3.I4.i4.p1.1.m1.1.2.2.2.cmml">C</mi><mo id="S3.I4.i4.p1.1.m1.1.2.2.1" xref="S3.I4.i4.p1.1.m1.1.2.2.1.cmml">⁢</mo><mrow id="S3.I4.i4.p1.1.m1.1.2.2.3.2" xref="S3.I4.i4.p1.1.m1.1.2.2.cmml"><mo id="S3.I4.i4.p1.1.m1.1.2.2.3.2.1" stretchy="false" xref="S3.I4.i4.p1.1.m1.1.2.2.cmml">(</mo><mi id="S3.I4.i4.p1.1.m1.1.1" xref="S3.I4.i4.p1.1.m1.1.1.cmml">D</mi><mo id="S3.I4.i4.p1.1.m1.1.2.2.3.2.2" stretchy="false" xref="S3.I4.i4.p1.1.m1.1.2.2.cmml">)</mo></mrow></mrow><mo id="S3.I4.i4.p1.1.m1.1.2.1" xref="S3.I4.i4.p1.1.m1.1.2.1.cmml"><</mo><mn id="S3.I4.i4.p1.1.m1.1.2.3" xref="S3.I4.i4.p1.1.m1.1.2.3.cmml">0.375</mn></mrow><annotation-xml encoding="MathML-Content" id="S3.I4.i4.p1.1.m1.1b"><apply id="S3.I4.i4.p1.1.m1.1.2.cmml" xref="S3.I4.i4.p1.1.m1.1.2"><lt id="S3.I4.i4.p1.1.m1.1.2.1.cmml" xref="S3.I4.i4.p1.1.m1.1.2.1"></lt><apply id="S3.I4.i4.p1.1.m1.1.2.2.cmml" xref="S3.I4.i4.p1.1.m1.1.2.2"><times id="S3.I4.i4.p1.1.m1.1.2.2.1.cmml" xref="S3.I4.i4.p1.1.m1.1.2.2.1"></times><ci id="S3.I4.i4.p1.1.m1.1.2.2.2.cmml" xref="S3.I4.i4.p1.1.m1.1.2.2.2">𝐶</ci><ci id="S3.I4.i4.p1.1.m1.1.1.cmml" xref="S3.I4.i4.p1.1.m1.1.1">𝐷</ci></apply><cn id="S3.I4.i4.p1.1.m1.1.2.3.cmml" type="float" xref="S3.I4.i4.p1.1.m1.1.2.3">0.375</cn></apply></annotation-xml><annotation encoding="application/x-tex" id="S3.I4.i4.p1.1.m1.1c">C(D)<0.375</annotation><annotation encoding="application/x-llamapun" id="S3.I4.i4.p1.1.m1.1d">italic_C ( italic_D ) < 0.375</annotation></semantics></math>): Uncertain prediction; flagged for additional verification.</p> </div> </li> </ul> </div> <div class="ltx_para ltx_noindent" id="S3.SS2.SSS3.p6"> <p class="ltx_p" id="S3.SS2.SSS3.p6.1">If no disease surpasses the confidence threshold <math alttext="\tau" class="ltx_Math" display="inline" id="S3.SS2.SSS3.p6.1.m1.1"><semantics id="S3.SS2.SSS3.p6.1.m1.1a"><mi id="S3.SS2.SSS3.p6.1.m1.1.1" xref="S3.SS2.SSS3.p6.1.m1.1.1.cmml">τ</mi><annotation-xml encoding="MathML-Content" id="S3.SS2.SSS3.p6.1.m1.1b"><ci id="S3.SS2.SSS3.p6.1.m1.1.1.cmml" xref="S3.SS2.SSS3.p6.1.m1.1.1">𝜏</ci></annotation-xml><annotation encoding="application/x-tex" id="S3.SS2.SSS3.p6.1.m1.1c">\tau</annotation><annotation encoding="application/x-llamapun" id="S3.SS2.SSS3.p6.1.m1.1d">italic_τ</annotation></semantics></math> (set dynamically, e.g., 0.375 or 0.75 depending on usage context), the case is flagged as Out-of-Distribution (OOD), triggering an escalation protocol. In such cases, the system either: (1) alerts a veterinary expert for manual review, or (2) recommends additional diagnostic tests (e.g., blood samples, pathogen screening) before reaching a final diagnosis.</p> </div> </section> </section> <section class="ltx_subsection" id="S3.SS3"> <h3 class="ltx_title ltx_title_subsection"> <span class="ltx_tag ltx_tag_subsection">3.3 </span>Recommendations</h3> <div class="ltx_para ltx_noindent" id="S3.SS3.p1"> <p class="ltx_p" id="S3.SS3.p1.1">The Recommendations module employs RAG techniques integrated with LLM-based reasoning to deliver context-aware, evidence-based veterinary guidance. This approach ensures that diagnostic outcomes are transformed into clear, actionable recommendations, providing users with precise and scientifically grounded advice.</p> </div> <section class="ltx_subsubsection" id="S3.SS3.SSS1"> <h4 class="ltx_title ltx_title_subsubsection"> <span class="ltx_tag ltx_tag_subsubsection">3.3.1 </span>Algorithmic Framework</h4> <div class="ltx_para ltx_noindent" id="S3.SS3.SSS1.p1"> <p class="ltx_p" id="S3.SS3.SSS1.p1.1">To ensure recommendations are context-aware and evidence-driven, the system employs a probabilistic approach, formulated as follows:</p> <table class="ltx_equation ltx_eqn_table" id="S3.Ex7"> <tbody><tr class="ltx_equation ltx_eqn_row ltx_align_baseline"> <td class="ltx_eqn_cell ltx_eqn_center_padleft"></td> <td class="ltx_eqn_cell ltx_align_center"><math alttext="P(R\mid D)=\sum_{z\in\mathcal{Z}}P(R\mid D,z)\,P(z\mid D)," class="ltx_Math" display="block" id="S3.Ex7.m1.3"><semantics id="S3.Ex7.m1.3a"><mrow id="S3.Ex7.m1.3.3.1" xref="S3.Ex7.m1.3.3.1.1.cmml"><mrow id="S3.Ex7.m1.3.3.1.1" xref="S3.Ex7.m1.3.3.1.1.cmml"><mrow id="S3.Ex7.m1.3.3.1.1.1" xref="S3.Ex7.m1.3.3.1.1.1.cmml"><mi id="S3.Ex7.m1.3.3.1.1.1.3" xref="S3.Ex7.m1.3.3.1.1.1.3.cmml">P</mi><mo id="S3.Ex7.m1.3.3.1.1.1.2" xref="S3.Ex7.m1.3.3.1.1.1.2.cmml">⁢</mo><mrow id="S3.Ex7.m1.3.3.1.1.1.1.1" xref="S3.Ex7.m1.3.3.1.1.1.1.1.1.cmml"><mo id="S3.Ex7.m1.3.3.1.1.1.1.1.2" stretchy="false" xref="S3.Ex7.m1.3.3.1.1.1.1.1.1.cmml">(</mo><mrow id="S3.Ex7.m1.3.3.1.1.1.1.1.1" xref="S3.Ex7.m1.3.3.1.1.1.1.1.1.cmml"><mi id="S3.Ex7.m1.3.3.1.1.1.1.1.1.2" xref="S3.Ex7.m1.3.3.1.1.1.1.1.1.2.cmml">R</mi><mo id="S3.Ex7.m1.3.3.1.1.1.1.1.1.1" xref="S3.Ex7.m1.3.3.1.1.1.1.1.1.1.cmml">∣</mo><mi id="S3.Ex7.m1.3.3.1.1.1.1.1.1.3" xref="S3.Ex7.m1.3.3.1.1.1.1.1.1.3.cmml">D</mi></mrow><mo id="S3.Ex7.m1.3.3.1.1.1.1.1.3" stretchy="false" xref="S3.Ex7.m1.3.3.1.1.1.1.1.1.cmml">)</mo></mrow></mrow><mo id="S3.Ex7.m1.3.3.1.1.4" rspace="0.111em" xref="S3.Ex7.m1.3.3.1.1.4.cmml">=</mo><mrow id="S3.Ex7.m1.3.3.1.1.3" xref="S3.Ex7.m1.3.3.1.1.3.cmml"><munder id="S3.Ex7.m1.3.3.1.1.3.3" xref="S3.Ex7.m1.3.3.1.1.3.3.cmml"><mo id="S3.Ex7.m1.3.3.1.1.3.3.2" movablelimits="false" xref="S3.Ex7.m1.3.3.1.1.3.3.2.cmml">∑</mo><mrow id="S3.Ex7.m1.3.3.1.1.3.3.3" xref="S3.Ex7.m1.3.3.1.1.3.3.3.cmml"><mi id="S3.Ex7.m1.3.3.1.1.3.3.3.2" xref="S3.Ex7.m1.3.3.1.1.3.3.3.2.cmml">z</mi><mo id="S3.Ex7.m1.3.3.1.1.3.3.3.1" xref="S3.Ex7.m1.3.3.1.1.3.3.3.1.cmml">∈</mo><mi class="ltx_font_mathcaligraphic" id="S3.Ex7.m1.3.3.1.1.3.3.3.3" xref="S3.Ex7.m1.3.3.1.1.3.3.3.3.cmml">𝒵</mi></mrow></munder><mrow id="S3.Ex7.m1.3.3.1.1.3.2" xref="S3.Ex7.m1.3.3.1.1.3.2.cmml"><mi id="S3.Ex7.m1.3.3.1.1.3.2.4" xref="S3.Ex7.m1.3.3.1.1.3.2.4.cmml">P</mi><mo id="S3.Ex7.m1.3.3.1.1.3.2.3" xref="S3.Ex7.m1.3.3.1.1.3.2.3.cmml">⁢</mo><mrow id="S3.Ex7.m1.3.3.1.1.2.1.1.1" xref="S3.Ex7.m1.3.3.1.1.2.1.1.1.1.cmml"><mo id="S3.Ex7.m1.3.3.1.1.2.1.1.1.2" stretchy="false" xref="S3.Ex7.m1.3.3.1.1.2.1.1.1.1.cmml">(</mo><mrow id="S3.Ex7.m1.3.3.1.1.2.1.1.1.1" xref="S3.Ex7.m1.3.3.1.1.2.1.1.1.1.cmml"><mi id="S3.Ex7.m1.3.3.1.1.2.1.1.1.1.2" xref="S3.Ex7.m1.3.3.1.1.2.1.1.1.1.2.cmml">R</mi><mo id="S3.Ex7.m1.3.3.1.1.2.1.1.1.1.1" xref="S3.Ex7.m1.3.3.1.1.2.1.1.1.1.1.cmml">∣</mo><mrow id="S3.Ex7.m1.3.3.1.1.2.1.1.1.1.3.2" xref="S3.Ex7.m1.3.3.1.1.2.1.1.1.1.3.1.cmml"><mi id="S3.Ex7.m1.1.1" xref="S3.Ex7.m1.1.1.cmml">D</mi><mo id="S3.Ex7.m1.3.3.1.1.2.1.1.1.1.3.2.1" xref="S3.Ex7.m1.3.3.1.1.2.1.1.1.1.3.1.cmml">,</mo><mi id="S3.Ex7.m1.2.2" xref="S3.Ex7.m1.2.2.cmml">z</mi></mrow></mrow><mo id="S3.Ex7.m1.3.3.1.1.2.1.1.1.3" stretchy="false" xref="S3.Ex7.m1.3.3.1.1.2.1.1.1.1.cmml">)</mo></mrow><mo id="S3.Ex7.m1.3.3.1.1.3.2.3a" lspace="0.170em" xref="S3.Ex7.m1.3.3.1.1.3.2.3.cmml">⁢</mo><mi id="S3.Ex7.m1.3.3.1.1.3.2.5" xref="S3.Ex7.m1.3.3.1.1.3.2.5.cmml">P</mi><mo id="S3.Ex7.m1.3.3.1.1.3.2.3b" xref="S3.Ex7.m1.3.3.1.1.3.2.3.cmml">⁢</mo><mrow id="S3.Ex7.m1.3.3.1.1.3.2.2.1" xref="S3.Ex7.m1.3.3.1.1.3.2.2.1.1.cmml"><mo id="S3.Ex7.m1.3.3.1.1.3.2.2.1.2" stretchy="false" xref="S3.Ex7.m1.3.3.1.1.3.2.2.1.1.cmml">(</mo><mrow id="S3.Ex7.m1.3.3.1.1.3.2.2.1.1" xref="S3.Ex7.m1.3.3.1.1.3.2.2.1.1.cmml"><mi id="S3.Ex7.m1.3.3.1.1.3.2.2.1.1.2" xref="S3.Ex7.m1.3.3.1.1.3.2.2.1.1.2.cmml">z</mi><mo id="S3.Ex7.m1.3.3.1.1.3.2.2.1.1.1" xref="S3.Ex7.m1.3.3.1.1.3.2.2.1.1.1.cmml">∣</mo><mi id="S3.Ex7.m1.3.3.1.1.3.2.2.1.1.3" xref="S3.Ex7.m1.3.3.1.1.3.2.2.1.1.3.cmml">D</mi></mrow><mo id="S3.Ex7.m1.3.3.1.1.3.2.2.1.3" stretchy="false" xref="S3.Ex7.m1.3.3.1.1.3.2.2.1.1.cmml">)</mo></mrow></mrow></mrow></mrow><mo id="S3.Ex7.m1.3.3.1.2" xref="S3.Ex7.m1.3.3.1.1.cmml">,</mo></mrow><annotation-xml encoding="MathML-Content" id="S3.Ex7.m1.3b"><apply id="S3.Ex7.m1.3.3.1.1.cmml" xref="S3.Ex7.m1.3.3.1"><eq id="S3.Ex7.m1.3.3.1.1.4.cmml" xref="S3.Ex7.m1.3.3.1.1.4"></eq><apply id="S3.Ex7.m1.3.3.1.1.1.cmml" xref="S3.Ex7.m1.3.3.1.1.1"><times id="S3.Ex7.m1.3.3.1.1.1.2.cmml" xref="S3.Ex7.m1.3.3.1.1.1.2"></times><ci id="S3.Ex7.m1.3.3.1.1.1.3.cmml" xref="S3.Ex7.m1.3.3.1.1.1.3">𝑃</ci><apply id="S3.Ex7.m1.3.3.1.1.1.1.1.1.cmml" xref="S3.Ex7.m1.3.3.1.1.1.1.1"><csymbol cd="latexml" id="S3.Ex7.m1.3.3.1.1.1.1.1.1.1.cmml" xref="S3.Ex7.m1.3.3.1.1.1.1.1.1.1">conditional</csymbol><ci id="S3.Ex7.m1.3.3.1.1.1.1.1.1.2.cmml" xref="S3.Ex7.m1.3.3.1.1.1.1.1.1.2">𝑅</ci><ci id="S3.Ex7.m1.3.3.1.1.1.1.1.1.3.cmml" xref="S3.Ex7.m1.3.3.1.1.1.1.1.1.3">𝐷</ci></apply></apply><apply id="S3.Ex7.m1.3.3.1.1.3.cmml" xref="S3.Ex7.m1.3.3.1.1.3"><apply id="S3.Ex7.m1.3.3.1.1.3.3.cmml" xref="S3.Ex7.m1.3.3.1.1.3.3"><csymbol cd="ambiguous" id="S3.Ex7.m1.3.3.1.1.3.3.1.cmml" xref="S3.Ex7.m1.3.3.1.1.3.3">subscript</csymbol><sum id="S3.Ex7.m1.3.3.1.1.3.3.2.cmml" xref="S3.Ex7.m1.3.3.1.1.3.3.2"></sum><apply id="S3.Ex7.m1.3.3.1.1.3.3.3.cmml" xref="S3.Ex7.m1.3.3.1.1.3.3.3"><in id="S3.Ex7.m1.3.3.1.1.3.3.3.1.cmml" xref="S3.Ex7.m1.3.3.1.1.3.3.3.1"></in><ci id="S3.Ex7.m1.3.3.1.1.3.3.3.2.cmml" xref="S3.Ex7.m1.3.3.1.1.3.3.3.2">𝑧</ci><ci id="S3.Ex7.m1.3.3.1.1.3.3.3.3.cmml" xref="S3.Ex7.m1.3.3.1.1.3.3.3.3">𝒵</ci></apply></apply><apply id="S3.Ex7.m1.3.3.1.1.3.2.cmml" xref="S3.Ex7.m1.3.3.1.1.3.2"><times id="S3.Ex7.m1.3.3.1.1.3.2.3.cmml" xref="S3.Ex7.m1.3.3.1.1.3.2.3"></times><ci id="S3.Ex7.m1.3.3.1.1.3.2.4.cmml" xref="S3.Ex7.m1.3.3.1.1.3.2.4">𝑃</ci><apply id="S3.Ex7.m1.3.3.1.1.2.1.1.1.1.cmml" xref="S3.Ex7.m1.3.3.1.1.2.1.1.1"><csymbol cd="latexml" id="S3.Ex7.m1.3.3.1.1.2.1.1.1.1.1.cmml" xref="S3.Ex7.m1.3.3.1.1.2.1.1.1.1.1">conditional</csymbol><ci id="S3.Ex7.m1.3.3.1.1.2.1.1.1.1.2.cmml" xref="S3.Ex7.m1.3.3.1.1.2.1.1.1.1.2">𝑅</ci><list id="S3.Ex7.m1.3.3.1.1.2.1.1.1.1.3.1.cmml" xref="S3.Ex7.m1.3.3.1.1.2.1.1.1.1.3.2"><ci id="S3.Ex7.m1.1.1.cmml" xref="S3.Ex7.m1.1.1">𝐷</ci><ci id="S3.Ex7.m1.2.2.cmml" xref="S3.Ex7.m1.2.2">𝑧</ci></list></apply><ci id="S3.Ex7.m1.3.3.1.1.3.2.5.cmml" xref="S3.Ex7.m1.3.3.1.1.3.2.5">𝑃</ci><apply id="S3.Ex7.m1.3.3.1.1.3.2.2.1.1.cmml" xref="S3.Ex7.m1.3.3.1.1.3.2.2.1"><csymbol cd="latexml" id="S3.Ex7.m1.3.3.1.1.3.2.2.1.1.1.cmml" xref="S3.Ex7.m1.3.3.1.1.3.2.2.1.1.1">conditional</csymbol><ci id="S3.Ex7.m1.3.3.1.1.3.2.2.1.1.2.cmml" xref="S3.Ex7.m1.3.3.1.1.3.2.2.1.1.2">𝑧</ci><ci id="S3.Ex7.m1.3.3.1.1.3.2.2.1.1.3.cmml" xref="S3.Ex7.m1.3.3.1.1.3.2.2.1.1.3">𝐷</ci></apply></apply></apply></apply></annotation-xml><annotation encoding="application/x-tex" id="S3.Ex7.m1.3c">P(R\mid D)=\sum_{z\in\mathcal{Z}}P(R\mid D,z)\,P(z\mid D),</annotation><annotation encoding="application/x-llamapun" id="S3.Ex7.m1.3d">italic_P ( italic_R ∣ italic_D ) = ∑ start_POSTSUBSCRIPT italic_z ∈ caligraphic_Z end_POSTSUBSCRIPT italic_P ( italic_R ∣ italic_D , italic_z ) italic_P ( italic_z ∣ italic_D ) ,</annotation></semantics></math></td> <td class="ltx_eqn_cell ltx_eqn_center_padright"></td> </tr></tbody> </table> <p class="ltx_p" id="S3.SS3.SSS1.p1.2">where:</p> <ul class="ltx_itemize" id="S3.I5"> <li class="ltx_item" id="S3.I5.i1" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item">•</span> <div class="ltx_para" id="S3.I5.i1.p1"> <p class="ltx_p" id="S3.I5.i1.p1.1"><math alttext="D" class="ltx_Math" display="inline" id="S3.I5.i1.p1.1.m1.1"><semantics id="S3.I5.i1.p1.1.m1.1a"><mi id="S3.I5.i1.p1.1.m1.1.1" xref="S3.I5.i1.p1.1.m1.1.1.cmml">D</mi><annotation-xml encoding="MathML-Content" id="S3.I5.i1.p1.1.m1.1b"><ci id="S3.I5.i1.p1.1.m1.1.1.cmml" xref="S3.I5.i1.p1.1.m1.1.1">𝐷</ci></annotation-xml><annotation encoding="application/x-tex" id="S3.I5.i1.p1.1.m1.1c">D</annotation><annotation encoding="application/x-llamapun" id="S3.I5.i1.p1.1.m1.1d">italic_D</annotation></semantics></math> denotes the diagnosed disease,</p> </div> </li> <li class="ltx_item" id="S3.I5.i2" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item">•</span> <div class="ltx_para" id="S3.I5.i2.p1"> <p class="ltx_p" id="S3.I5.i2.p1.1"><math alttext="R" class="ltx_Math" display="inline" id="S3.I5.i2.p1.1.m1.1"><semantics id="S3.I5.i2.p1.1.m1.1a"><mi id="S3.I5.i2.p1.1.m1.1.1" xref="S3.I5.i2.p1.1.m1.1.1.cmml">R</mi><annotation-xml encoding="MathML-Content" id="S3.I5.i2.p1.1.m1.1b"><ci id="S3.I5.i2.p1.1.m1.1.1.cmml" xref="S3.I5.i2.p1.1.m1.1.1">𝑅</ci></annotation-xml><annotation encoding="application/x-tex" id="S3.I5.i2.p1.1.m1.1c">R</annotation><annotation encoding="application/x-llamapun" id="S3.I5.i2.p1.1.m1.1d">italic_R</annotation></semantics></math> represents the generated recommendation,</p> </div> </li> <li class="ltx_item" id="S3.I5.i3" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item">•</span> <div class="ltx_para" id="S3.I5.i3.p1"> <p class="ltx_p" id="S3.I5.i3.p1.2"><math alttext="z" class="ltx_Math" display="inline" id="S3.I5.i3.p1.1.m1.1"><semantics id="S3.I5.i3.p1.1.m1.1a"><mi id="S3.I5.i3.p1.1.m1.1.1" xref="S3.I5.i3.p1.1.m1.1.1.cmml">z</mi><annotation-xml encoding="MathML-Content" id="S3.I5.i3.p1.1.m1.1b"><ci id="S3.I5.i3.p1.1.m1.1.1.cmml" xref="S3.I5.i3.p1.1.m1.1.1">𝑧</ci></annotation-xml><annotation encoding="application/x-tex" id="S3.I5.i3.p1.1.m1.1c">z</annotation><annotation encoding="application/x-llamapun" id="S3.I5.i3.p1.1.m1.1d">italic_z</annotation></semantics></math> is a document retrieved from the veterinary knowledge base <math alttext="\mathcal{Z}" class="ltx_Math" display="inline" id="S3.I5.i3.p1.2.m2.1"><semantics id="S3.I5.i3.p1.2.m2.1a"><mi class="ltx_font_mathcaligraphic" id="S3.I5.i3.p1.2.m2.1.1" xref="S3.I5.i3.p1.2.m2.1.1.cmml">𝒵</mi><annotation-xml encoding="MathML-Content" id="S3.I5.i3.p1.2.m2.1b"><ci id="S3.I5.i3.p1.2.m2.1.1.cmml" xref="S3.I5.i3.p1.2.m2.1.1">𝒵</ci></annotation-xml><annotation encoding="application/x-tex" id="S3.I5.i3.p1.2.m2.1c">\mathcal{Z}</annotation><annotation encoding="application/x-llamapun" id="S3.I5.i3.p1.2.m2.1d">caligraphic_Z</annotation></semantics></math>,</p> </div> </li> <li class="ltx_item" id="S3.I5.i4" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item">•</span> <div class="ltx_para" id="S3.I5.i4.p1"> <p class="ltx_p" id="S3.I5.i4.p1.3"><math alttext="P(z\mid D)" class="ltx_Math" display="inline" id="S3.I5.i4.p1.1.m1.1"><semantics id="S3.I5.i4.p1.1.m1.1a"><mrow id="S3.I5.i4.p1.1.m1.1.1" xref="S3.I5.i4.p1.1.m1.1.1.cmml"><mi id="S3.I5.i4.p1.1.m1.1.1.3" xref="S3.I5.i4.p1.1.m1.1.1.3.cmml">P</mi><mo id="S3.I5.i4.p1.1.m1.1.1.2" xref="S3.I5.i4.p1.1.m1.1.1.2.cmml">⁢</mo><mrow id="S3.I5.i4.p1.1.m1.1.1.1.1" xref="S3.I5.i4.p1.1.m1.1.1.1.1.1.cmml"><mo id="S3.I5.i4.p1.1.m1.1.1.1.1.2" stretchy="false" xref="S3.I5.i4.p1.1.m1.1.1.1.1.1.cmml">(</mo><mrow id="S3.I5.i4.p1.1.m1.1.1.1.1.1" xref="S3.I5.i4.p1.1.m1.1.1.1.1.1.cmml"><mi id="S3.I5.i4.p1.1.m1.1.1.1.1.1.2" xref="S3.I5.i4.p1.1.m1.1.1.1.1.1.2.cmml">z</mi><mo id="S3.I5.i4.p1.1.m1.1.1.1.1.1.1" xref="S3.I5.i4.p1.1.m1.1.1.1.1.1.1.cmml">∣</mo><mi id="S3.I5.i4.p1.1.m1.1.1.1.1.1.3" xref="S3.I5.i4.p1.1.m1.1.1.1.1.1.3.cmml">D</mi></mrow><mo id="S3.I5.i4.p1.1.m1.1.1.1.1.3" stretchy="false" xref="S3.I5.i4.p1.1.m1.1.1.1.1.1.cmml">)</mo></mrow></mrow><annotation-xml encoding="MathML-Content" id="S3.I5.i4.p1.1.m1.1b"><apply id="S3.I5.i4.p1.1.m1.1.1.cmml" xref="S3.I5.i4.p1.1.m1.1.1"><times id="S3.I5.i4.p1.1.m1.1.1.2.cmml" xref="S3.I5.i4.p1.1.m1.1.1.2"></times><ci id="S3.I5.i4.p1.1.m1.1.1.3.cmml" xref="S3.I5.i4.p1.1.m1.1.1.3">𝑃</ci><apply id="S3.I5.i4.p1.1.m1.1.1.1.1.1.cmml" xref="S3.I5.i4.p1.1.m1.1.1.1.1"><csymbol cd="latexml" id="S3.I5.i4.p1.1.m1.1.1.1.1.1.1.cmml" xref="S3.I5.i4.p1.1.m1.1.1.1.1.1.1">conditional</csymbol><ci id="S3.I5.i4.p1.1.m1.1.1.1.1.1.2.cmml" xref="S3.I5.i4.p1.1.m1.1.1.1.1.1.2">𝑧</ci><ci id="S3.I5.i4.p1.1.m1.1.1.1.1.1.3.cmml" xref="S3.I5.i4.p1.1.m1.1.1.1.1.1.3">𝐷</ci></apply></apply></annotation-xml><annotation encoding="application/x-tex" id="S3.I5.i4.p1.1.m1.1c">P(z\mid D)</annotation><annotation encoding="application/x-llamapun" id="S3.I5.i4.p1.1.m1.1d">italic_P ( italic_z ∣ italic_D )</annotation></semantics></math> represents the probability of retrieving document <math alttext="z" class="ltx_Math" display="inline" id="S3.I5.i4.p1.2.m2.1"><semantics id="S3.I5.i4.p1.2.m2.1a"><mi id="S3.I5.i4.p1.2.m2.1.1" xref="S3.I5.i4.p1.2.m2.1.1.cmml">z</mi><annotation-xml encoding="MathML-Content" id="S3.I5.i4.p1.2.m2.1b"><ci id="S3.I5.i4.p1.2.m2.1.1.cmml" xref="S3.I5.i4.p1.2.m2.1.1">𝑧</ci></annotation-xml><annotation encoding="application/x-tex" id="S3.I5.i4.p1.2.m2.1c">z</annotation><annotation encoding="application/x-llamapun" id="S3.I5.i4.p1.2.m2.1d">italic_z</annotation></semantics></math> given <math alttext="D" class="ltx_Math" display="inline" id="S3.I5.i4.p1.3.m3.1"><semantics id="S3.I5.i4.p1.3.m3.1a"><mi id="S3.I5.i4.p1.3.m3.1.1" xref="S3.I5.i4.p1.3.m3.1.1.cmml">D</mi><annotation-xml encoding="MathML-Content" id="S3.I5.i4.p1.3.m3.1b"><ci id="S3.I5.i4.p1.3.m3.1.1.cmml" xref="S3.I5.i4.p1.3.m3.1.1">𝐷</ci></annotation-xml><annotation encoding="application/x-tex" id="S3.I5.i4.p1.3.m3.1c">D</annotation><annotation encoding="application/x-llamapun" id="S3.I5.i4.p1.3.m3.1d">italic_D</annotation></semantics></math>,</p> </div> </li> <li class="ltx_item" id="S3.I5.i5" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item">•</span> <div class="ltx_para ltx_noindent" id="S3.I5.i5.p1"> <p class="ltx_p" id="S3.I5.i5.p1.4"><math alttext="P(R\mid D,z)" class="ltx_Math" display="inline" id="S3.I5.i5.p1.1.m1.3"><semantics id="S3.I5.i5.p1.1.m1.3a"><mrow id="S3.I5.i5.p1.1.m1.3.3" xref="S3.I5.i5.p1.1.m1.3.3.cmml"><mi id="S3.I5.i5.p1.1.m1.3.3.3" xref="S3.I5.i5.p1.1.m1.3.3.3.cmml">P</mi><mo id="S3.I5.i5.p1.1.m1.3.3.2" xref="S3.I5.i5.p1.1.m1.3.3.2.cmml">⁢</mo><mrow id="S3.I5.i5.p1.1.m1.3.3.1.1" xref="S3.I5.i5.p1.1.m1.3.3.1.1.1.cmml"><mo id="S3.I5.i5.p1.1.m1.3.3.1.1.2" stretchy="false" xref="S3.I5.i5.p1.1.m1.3.3.1.1.1.cmml">(</mo><mrow id="S3.I5.i5.p1.1.m1.3.3.1.1.1" xref="S3.I5.i5.p1.1.m1.3.3.1.1.1.cmml"><mi id="S3.I5.i5.p1.1.m1.3.3.1.1.1.2" xref="S3.I5.i5.p1.1.m1.3.3.1.1.1.2.cmml">R</mi><mo id="S3.I5.i5.p1.1.m1.3.3.1.1.1.1" xref="S3.I5.i5.p1.1.m1.3.3.1.1.1.1.cmml">∣</mo><mrow id="S3.I5.i5.p1.1.m1.3.3.1.1.1.3.2" xref="S3.I5.i5.p1.1.m1.3.3.1.1.1.3.1.cmml"><mi id="S3.I5.i5.p1.1.m1.1.1" xref="S3.I5.i5.p1.1.m1.1.1.cmml">D</mi><mo id="S3.I5.i5.p1.1.m1.3.3.1.1.1.3.2.1" xref="S3.I5.i5.p1.1.m1.3.3.1.1.1.3.1.cmml">,</mo><mi id="S3.I5.i5.p1.1.m1.2.2" xref="S3.I5.i5.p1.1.m1.2.2.cmml">z</mi></mrow></mrow><mo id="S3.I5.i5.p1.1.m1.3.3.1.1.3" stretchy="false" xref="S3.I5.i5.p1.1.m1.3.3.1.1.1.cmml">)</mo></mrow></mrow><annotation-xml encoding="MathML-Content" id="S3.I5.i5.p1.1.m1.3b"><apply id="S3.I5.i5.p1.1.m1.3.3.cmml" xref="S3.I5.i5.p1.1.m1.3.3"><times id="S3.I5.i5.p1.1.m1.3.3.2.cmml" xref="S3.I5.i5.p1.1.m1.3.3.2"></times><ci id="S3.I5.i5.p1.1.m1.3.3.3.cmml" xref="S3.I5.i5.p1.1.m1.3.3.3">𝑃</ci><apply id="S3.I5.i5.p1.1.m1.3.3.1.1.1.cmml" xref="S3.I5.i5.p1.1.m1.3.3.1.1"><csymbol cd="latexml" id="S3.I5.i5.p1.1.m1.3.3.1.1.1.1.cmml" xref="S3.I5.i5.p1.1.m1.3.3.1.1.1.1">conditional</csymbol><ci id="S3.I5.i5.p1.1.m1.3.3.1.1.1.2.cmml" xref="S3.I5.i5.p1.1.m1.3.3.1.1.1.2">𝑅</ci><list id="S3.I5.i5.p1.1.m1.3.3.1.1.1.3.1.cmml" xref="S3.I5.i5.p1.1.m1.3.3.1.1.1.3.2"><ci id="S3.I5.i5.p1.1.m1.1.1.cmml" xref="S3.I5.i5.p1.1.m1.1.1">𝐷</ci><ci id="S3.I5.i5.p1.1.m1.2.2.cmml" xref="S3.I5.i5.p1.1.m1.2.2">𝑧</ci></list></apply></apply></annotation-xml><annotation encoding="application/x-tex" id="S3.I5.i5.p1.1.m1.3c">P(R\mid D,z)</annotation><annotation encoding="application/x-llamapun" id="S3.I5.i5.p1.1.m1.3d">italic_P ( italic_R ∣ italic_D , italic_z )</annotation></semantics></math> denotes the probability of generating <math alttext="R" class="ltx_Math" display="inline" id="S3.I5.i5.p1.2.m2.1"><semantics id="S3.I5.i5.p1.2.m2.1a"><mi id="S3.I5.i5.p1.2.m2.1.1" xref="S3.I5.i5.p1.2.m2.1.1.cmml">R</mi><annotation-xml encoding="MathML-Content" id="S3.I5.i5.p1.2.m2.1b"><ci id="S3.I5.i5.p1.2.m2.1.1.cmml" xref="S3.I5.i5.p1.2.m2.1.1">𝑅</ci></annotation-xml><annotation encoding="application/x-tex" id="S3.I5.i5.p1.2.m2.1c">R</annotation><annotation encoding="application/x-llamapun" id="S3.I5.i5.p1.2.m2.1d">italic_R</annotation></semantics></math> conditioned on both <math alttext="D" class="ltx_Math" display="inline" id="S3.I5.i5.p1.3.m3.1"><semantics id="S3.I5.i5.p1.3.m3.1a"><mi id="S3.I5.i5.p1.3.m3.1.1" xref="S3.I5.i5.p1.3.m3.1.1.cmml">D</mi><annotation-xml encoding="MathML-Content" id="S3.I5.i5.p1.3.m3.1b"><ci id="S3.I5.i5.p1.3.m3.1.1.cmml" xref="S3.I5.i5.p1.3.m3.1.1">𝐷</ci></annotation-xml><annotation encoding="application/x-tex" id="S3.I5.i5.p1.3.m3.1c">D</annotation><annotation encoding="application/x-llamapun" id="S3.I5.i5.p1.3.m3.1d">italic_D</annotation></semantics></math> and <math alttext="z" class="ltx_Math" display="inline" id="S3.I5.i5.p1.4.m4.1"><semantics id="S3.I5.i5.p1.4.m4.1a"><mi id="S3.I5.i5.p1.4.m4.1.1" xref="S3.I5.i5.p1.4.m4.1.1.cmml">z</mi><annotation-xml encoding="MathML-Content" id="S3.I5.i5.p1.4.m4.1b"><ci id="S3.I5.i5.p1.4.m4.1.1.cmml" xref="S3.I5.i5.p1.4.m4.1.1">𝑧</ci></annotation-xml><annotation encoding="application/x-tex" id="S3.I5.i5.p1.4.m4.1c">z</annotation><annotation encoding="application/x-llamapun" id="S3.I5.i5.p1.4.m4.1d">italic_z</annotation></semantics></math>.</p> </div> </li> </ul> </div> <div class="ltx_para ltx_noindent" id="S3.SS3.SSS1.p2"> <p class="ltx_p" id="S3.SS3.SSS1.p2.1">To ensure robust performance, an exponential backoff strategy with five steps is integrated into the LLM-based generation process. If an initial LLM call does not yield a satisfactory result, the system automatically retries the request with progressively longer delay intervals. This iterative process continues until <math alttext="P(R\mid D,z)" class="ltx_Math" display="inline" id="S3.SS3.SSS1.p2.1.m1.3"><semantics id="S3.SS3.SSS1.p2.1.m1.3a"><mrow id="S3.SS3.SSS1.p2.1.m1.3.3" xref="S3.SS3.SSS1.p2.1.m1.3.3.cmml"><mi id="S3.SS3.SSS1.p2.1.m1.3.3.3" xref="S3.SS3.SSS1.p2.1.m1.3.3.3.cmml">P</mi><mo id="S3.SS3.SSS1.p2.1.m1.3.3.2" xref="S3.SS3.SSS1.p2.1.m1.3.3.2.cmml">⁢</mo><mrow id="S3.SS3.SSS1.p2.1.m1.3.3.1.1" xref="S3.SS3.SSS1.p2.1.m1.3.3.1.1.1.cmml"><mo id="S3.SS3.SSS1.p2.1.m1.3.3.1.1.2" stretchy="false" xref="S3.SS3.SSS1.p2.1.m1.3.3.1.1.1.cmml">(</mo><mrow id="S3.SS3.SSS1.p2.1.m1.3.3.1.1.1" xref="S3.SS3.SSS1.p2.1.m1.3.3.1.1.1.cmml"><mi id="S3.SS3.SSS1.p2.1.m1.3.3.1.1.1.2" xref="S3.SS3.SSS1.p2.1.m1.3.3.1.1.1.2.cmml">R</mi><mo id="S3.SS3.SSS1.p2.1.m1.3.3.1.1.1.1" xref="S3.SS3.SSS1.p2.1.m1.3.3.1.1.1.1.cmml">∣</mo><mrow id="S3.SS3.SSS1.p2.1.m1.3.3.1.1.1.3.2" xref="S3.SS3.SSS1.p2.1.m1.3.3.1.1.1.3.1.cmml"><mi id="S3.SS3.SSS1.p2.1.m1.1.1" xref="S3.SS3.SSS1.p2.1.m1.1.1.cmml">D</mi><mo id="S3.SS3.SSS1.p2.1.m1.3.3.1.1.1.3.2.1" xref="S3.SS3.SSS1.p2.1.m1.3.3.1.1.1.3.1.cmml">,</mo><mi id="S3.SS3.SSS1.p2.1.m1.2.2" xref="S3.SS3.SSS1.p2.1.m1.2.2.cmml">z</mi></mrow></mrow><mo id="S3.SS3.SSS1.p2.1.m1.3.3.1.1.3" stretchy="false" xref="S3.SS3.SSS1.p2.1.m1.3.3.1.1.1.cmml">)</mo></mrow></mrow><annotation-xml encoding="MathML-Content" id="S3.SS3.SSS1.p2.1.m1.3b"><apply id="S3.SS3.SSS1.p2.1.m1.3.3.cmml" xref="S3.SS3.SSS1.p2.1.m1.3.3"><times id="S3.SS3.SSS1.p2.1.m1.3.3.2.cmml" xref="S3.SS3.SSS1.p2.1.m1.3.3.2"></times><ci id="S3.SS3.SSS1.p2.1.m1.3.3.3.cmml" xref="S3.SS3.SSS1.p2.1.m1.3.3.3">𝑃</ci><apply id="S3.SS3.SSS1.p2.1.m1.3.3.1.1.1.cmml" xref="S3.SS3.SSS1.p2.1.m1.3.3.1.1"><csymbol cd="latexml" id="S3.SS3.SSS1.p2.1.m1.3.3.1.1.1.1.cmml" xref="S3.SS3.SSS1.p2.1.m1.3.3.1.1.1.1">conditional</csymbol><ci id="S3.SS3.SSS1.p2.1.m1.3.3.1.1.1.2.cmml" xref="S3.SS3.SSS1.p2.1.m1.3.3.1.1.1.2">𝑅</ci><list id="S3.SS3.SSS1.p2.1.m1.3.3.1.1.1.3.1.cmml" xref="S3.SS3.SSS1.p2.1.m1.3.3.1.1.1.3.2"><ci id="S3.SS3.SSS1.p2.1.m1.1.1.cmml" xref="S3.SS3.SSS1.p2.1.m1.1.1">𝐷</ci><ci id="S3.SS3.SSS1.p2.1.m1.2.2.cmml" xref="S3.SS3.SSS1.p2.1.m1.2.2">𝑧</ci></list></apply></apply></annotation-xml><annotation encoding="application/x-tex" id="S3.SS3.SSS1.p2.1.m1.3c">P(R\mid D,z)</annotation><annotation encoding="application/x-llamapun" id="S3.SS3.SSS1.p2.1.m1.3d">italic_P ( italic_R ∣ italic_D , italic_z )</annotation></semantics></math> is sufficiently refined to produce a reliable and actionable recommendation.</p> </div> </section> <section class="ltx_subsubsection" id="S3.SS3.SSS2"> <h4 class="ltx_title ltx_title_subsubsection"> <span class="ltx_tag ltx_tag_subsubsection">3.3.2 </span>Integrated Recommendation Pipeline</h4> <div class="ltx_para ltx_noindent" id="S3.SS3.SSS2.p1"> <p class="ltx_p" id="S3.SS3.SSS2.p1.1">Building on the algorithmic foundation described above, the pipeline is structured into several key components, each designed to translate theoretical principles into practical steps:</p> </div> <div class="ltx_para ltx_noindent" id="S3.SS3.SSS2.p2"> <ul class="ltx_itemize" id="S3.I6"> <li class="ltx_item" id="S3.I6.i1" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item">•</span> <div class="ltx_para ltx_noindent" id="S3.I6.i1.p1"> <p class="ltx_p" id="S3.I6.i1.p1.1"><span class="ltx_text ltx_font_bold" id="S3.I6.i1.p1.1.1">Registration State:</span> Upon receiving a diagnosis <math alttext="D" class="ltx_Math" display="inline" id="S3.I6.i1.p1.1.m1.1"><semantics id="S3.I6.i1.p1.1.m1.1a"><mi id="S3.I6.i1.p1.1.m1.1.1" xref="S3.I6.i1.p1.1.m1.1.1.cmml">D</mi><annotation-xml encoding="MathML-Content" id="S3.I6.i1.p1.1.m1.1b"><ci id="S3.I6.i1.p1.1.m1.1.1.cmml" xref="S3.I6.i1.p1.1.m1.1.1">𝐷</ci></annotation-xml><annotation encoding="application/x-tex" id="S3.I6.i1.p1.1.m1.1c">D</annotation><annotation encoding="application/x-llamapun" id="S3.I6.i1.p1.1.m1.1d">italic_D</annotation></semantics></math> along with the corresponding user query, the system establishes a task-specific state that encapsulates query details, conversation history, and relevant metadata. Concurrently, a tailored filter expression is formulated to guide the document retrieval process.</p> </div> </li> <li class="ltx_item" id="S3.I6.i2" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item">•</span> <div class="ltx_para ltx_noindent" id="S3.I6.i2.p1"> <p class="ltx_p" id="S3.I6.i2.p1.1"><span class="ltx_text ltx_font_bold" id="S3.I6.i2.p1.1.1">Entity Extraction:</span> Two complementary retrieval mechanisms are employed:</p> <ul class="ltx_itemize" id="S3.I6.i2.I1"> <li class="ltx_item" id="S3.I6.i2.I1.i1" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item"><span class="ltx_text ltx_font_bold" id="S3.I6.i2.I1.i1.1.1.1">–</span></span> <div class="ltx_para" id="S3.I6.i2.I1.i1.p1"> <p class="ltx_p" id="S3.I6.i2.I1.i1.p1.1">General Entity Extraction: Identifies key veterinary terms and domain-specific terminology used by experts. This includes specialized disease references, such as “Roll Over,” which describes a swine health status transition from disease-free to infected. These entities are retrieved from the knowledge base relevant to <math alttext="D" class="ltx_Math" display="inline" id="S3.I6.i2.I1.i1.p1.1.m1.1"><semantics id="S3.I6.i2.I1.i1.p1.1.m1.1a"><mi id="S3.I6.i2.I1.i1.p1.1.m1.1.1" xref="S3.I6.i2.I1.i1.p1.1.m1.1.1.cmml">D</mi><annotation-xml encoding="MathML-Content" id="S3.I6.i2.I1.i1.p1.1.m1.1b"><ci id="S3.I6.i2.I1.i1.p1.1.m1.1.1.cmml" xref="S3.I6.i2.I1.i1.p1.1.m1.1.1">𝐷</ci></annotation-xml><annotation encoding="application/x-tex" id="S3.I6.i2.I1.i1.p1.1.m1.1c">D</annotation><annotation encoding="application/x-llamapun" id="S3.I6.i2.I1.i1.p1.1.m1.1d">italic_D</annotation></semantics></math> to ensure accurate context.</p> </div> </li> <li class="ltx_item" id="S3.I6.i2.I1.i2" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item"><span class="ltx_text ltx_font_bold" id="S3.I6.i2.I1.i2.1.1.1">–</span></span> <div class="ltx_para ltx_noindent" id="S3.I6.i2.I1.i2.p1"> <p class="ltx_p" id="S3.I6.i2.I1.i2.p1.1">Medicine and Vaccine-Specific Extraction: Focuses on isolating pharmaceutical and vaccine-related information, such as trade names and medicine or vaccine groups, ensuring precise medical guidance in recommendations.</p> </div> </li> </ul> </div> </li> <li class="ltx_item" id="S3.I6.i3" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item">•</span> <div class="ltx_para ltx_noindent" id="S3.I6.i3.p1"> <p class="ltx_p" id="S3.I6.i3.p1.1"><span class="ltx_text ltx_font_bold" id="S3.I6.i3.p1.1.1">Query Contextualization:</span> A two-stage rewriting process refines the original query. First, general entities and conversation history are used to construct a detailed query. Then, vaccine-specific information is integrated, enhancing query precision.</p> </div> </li> <li class="ltx_item" id="S3.I6.i4" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item">•</span> <div class="ltx_para ltx_noindent" id="S3.I6.i4.p1"> <p class="ltx_p" id="S3.I6.i4.p1.1"><span class="ltx_text ltx_font_bold" id="S3.I6.i4.p1.1.1">Document Retrieval and Processing:</span> The refined query is used to retrieve documents from a vectorized data store. Extracted documents, along with their metadata, are consolidated to establish a robust context for the final reasoning phase.</p> </div> </li> <li class="ltx_item" id="S3.I6.i5" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item">•</span> <div class="ltx_para ltx_noindent" id="S3.I6.i5.p1"> <p class="ltx_p" id="S3.I6.i5.p1.1"><span class="ltx_text ltx_font_bold" id="S3.I6.i5.p1.1.1">LLM-Based Answer Generation:</span> The comprehensive context—augmented with historical interactions and other relevant parameters—is provided as input to an LLM. An exponential backoff strategy ensures resilient and reliable LLM call management.</p> </div> </li> <li class="ltx_item" id="S3.I6.i6" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item">•</span> <div class="ltx_para ltx_noindent" id="S3.I6.i6.p1"> <p class="ltx_p" id="S3.I6.i6.p1.1"><span class="ltx_text ltx_font_bold" id="S3.I6.i6.p1.1.1">Output Assembly:</span> Finally, the system aggregates the generated recommendation along with retrieved documents and associated metadata. The final output is presented to the user in a clear and actionable format.</p> </div> </li> </ul> </div> </section> </section> </section> <section class="ltx_section" id="S4"> <h2 class="ltx_title ltx_title_section"> <span class="ltx_tag ltx_tag_section">4 </span>Experimental Setup</h2> <div class="ltx_para ltx_noindent" id="S4.p1"> <p class="ltx_p" id="S4.p1.1">The AI system’s performance is evaluated across three tasks: Query Classification, Disease Diagnosis, and Knowledge Retrieval. These tasks assess its ability to classify queries, diagnose diseases accurately, and retrieve relevant medical knowledge. The evaluation framework ensures precision and reliability by measuring accuracy, effectiveness, and relevance. The experimental design separates training, validation, and test datasets to evaluate both internal consistency and external generalizability.</p> </div> <section class="ltx_subsection" id="S4.SS1"> <h3 class="ltx_title ltx_title_subsection"> <span class="ltx_tag ltx_tag_subsection">4.1 </span>Data Collection</h3> <div class="ltx_para ltx_noindent" id="S4.SS1.p1"> <p class="ltx_p" id="S4.SS1.p1.1">To maintain consistency and facilitate precise comparisons across tasks, dataset details are summarized in a single table (Table <a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#S4.T1" title="Table 1 ‣ 4.1 Data Collection ‣ 4 Experimental Setup ‣ When Pigs Get Sick: Multi-Agent AI for Swine Disease Detection"><span class="ltx_text ltx_ref_tag">1</span></a>). This table provides a structured summary of the distribution of validation and test questions for each task, ensuring clarity and uniformity in data representation.</p> </div> <figure class="ltx_table" id="S4.T1"> <table class="ltx_tabular ltx_centering ltx_guessed_headers ltx_align_middle" id="S4.T1.1"> <thead class="ltx_thead"> <tr class="ltx_tr" id="S4.T1.1.1.1"> <th class="ltx_td ltx_align_left ltx_th ltx_th_column ltx_th_row ltx_border_l ltx_border_r ltx_border_t" id="S4.T1.1.1.1.1" style="padding-top:2.5pt;padding-bottom:2.5pt;"><span class="ltx_text ltx_font_bold" id="S4.T1.1.1.1.1.1">Task</span></th> <th class="ltx_td ltx_align_center ltx_th ltx_th_column ltx_border_r ltx_border_t" id="S4.T1.1.1.1.2" style="padding-top:2.5pt;padding-bottom:2.5pt;"><span class="ltx_text ltx_font_bold" id="S4.T1.1.1.1.2.1">Validation Questions</span></th> <th class="ltx_td ltx_align_center ltx_th ltx_th_column ltx_border_r ltx_border_t" id="S4.T1.1.1.1.3" style="padding-top:2.5pt;padding-bottom:2.5pt;"><span class="ltx_text ltx_font_bold" id="S4.T1.1.1.1.3.1">Test Questions</span></th> </tr> </thead> <tbody class="ltx_tbody"> <tr class="ltx_tr" id="S4.T1.1.2.1"> <th class="ltx_td ltx_align_left ltx_th ltx_th_row ltx_border_l ltx_border_r ltx_border_t" id="S4.T1.1.2.1.1" style="padding-top:2.5pt;padding-bottom:2.5pt;">Query Classification</th> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S4.T1.1.2.1.2" style="padding-top:2.5pt;padding-bottom:2.5pt;">220</td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S4.T1.1.2.1.3" style="padding-top:2.5pt;padding-bottom:2.5pt;">461</td> </tr> <tr class="ltx_tr" id="S4.T1.1.3.2"> <th class="ltx_td ltx_align_left ltx_th ltx_th_row ltx_border_l ltx_border_r" id="S4.T1.1.3.2.1" style="padding-top:2.5pt;padding-bottom:2.5pt;">Disease Diagnosis</th> <td class="ltx_td ltx_align_center ltx_border_r" id="S4.T1.1.3.2.2" style="padding-top:2.5pt;padding-bottom:2.5pt;">52</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S4.T1.1.3.2.3" style="padding-top:2.5pt;padding-bottom:2.5pt;">32</td> </tr> <tr class="ltx_tr" id="S4.T1.1.4.3"> <th class="ltx_td ltx_align_left ltx_th ltx_th_row ltx_border_l ltx_border_r" id="S4.T1.1.4.3.1" style="padding-top:2.5pt;padding-bottom:2.5pt;">Disease-Related Knowledge Retrieval</th> <td class="ltx_td ltx_align_center ltx_border_r" id="S4.T1.1.4.3.2" style="padding-top:2.5pt;padding-bottom:2.5pt;">2,808</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S4.T1.1.4.3.3" style="padding-top:2.5pt;padding-bottom:2.5pt;">128</td> </tr> <tr class="ltx_tr" id="S4.T1.1.5.4"> <th class="ltx_td ltx_align_left ltx_th ltx_th_row ltx_border_l ltx_border_r" id="S4.T1.1.5.4.1" style="padding-top:2.5pt;padding-bottom:2.5pt;">Vaccine-Related Knowledge Retrieval</th> <td class="ltx_td ltx_align_center ltx_border_r" id="S4.T1.1.5.4.2" style="padding-top:2.5pt;padding-bottom:2.5pt;">906</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S4.T1.1.5.4.3" style="padding-top:2.5pt;padding-bottom:2.5pt;">191</td> </tr> <tr class="ltx_tr" id="S4.T1.1.6.5" style="background-color:#E6E6E6;"> <th class="ltx_td ltx_align_left ltx_th ltx_th_row ltx_border_b ltx_border_l ltx_border_r ltx_border_t" id="S4.T1.1.6.5.1" style="padding-top:2.5pt;padding-bottom:2.5pt;"><span class="ltx_text ltx_font_bold" id="S4.T1.1.6.5.1.1" style="background-color:#E6E6E6;">Total</span></th> <td class="ltx_td ltx_align_center ltx_border_b ltx_border_r ltx_border_t" id="S4.T1.1.6.5.2" style="padding-top:2.5pt;padding-bottom:2.5pt;"><span class="ltx_text ltx_font_bold" id="S4.T1.1.6.5.2.1" style="background-color:#E6E6E6;">3,986</span></td> <td class="ltx_td ltx_align_center ltx_border_b ltx_border_r ltx_border_t" id="S4.T1.1.6.5.3" style="padding-top:2.5pt;padding-bottom:2.5pt;"><span class="ltx_text ltx_font_bold" id="S4.T1.1.6.5.3.1" style="background-color:#E6E6E6;">812</span></td> </tr> </tbody> </table> <figcaption class="ltx_caption ltx_centering"><span class="ltx_tag ltx_tag_table">Table 1: </span>Comprehensive dataset summary detailing the total number of validation and test questions across tasks.</figcaption> </figure> <div class="ltx_para ltx_noindent" id="S4.SS1.p2"> <p class="ltx_p" id="S4.SS1.p2.1">Both the validation and test sets are constructed using the same methodology, with the primary distinction being their source of origin:</p> <ul class="ltx_itemize" id="S4.I1"> <li class="ltx_item" id="S4.I1.i1" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item">•</span> <div class="ltx_para" id="S4.I1.i1.p1"> <p class="ltx_p" id="S4.I1.i1.p1.1"><span class="ltx_text ltx_font_bold" id="S4.I1.i1.p1.1.1">Validation Set:</span> AI-generated following structured guidelines.</p> </div> </li> <li class="ltx_item" id="S4.I1.i2" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item">•</span> <div class="ltx_para ltx_noindent" id="S4.I1.i2.p1"> <p class="ltx_p" id="S4.I1.i2.p1.1"><span class="ltx_text ltx_font_bold" id="S4.I1.i2.p1.1.1">Test Set:</span> Curated by veterinary experts.</p> </div> </li> </ul> </div> <div class="ltx_para ltx_noindent" id="S4.SS1.p3"> <p class="ltx_p" id="S4.SS1.p3.1">The questions were systematically derived from relevant references, including:</p> <ul class="ltx_itemize" id="S4.I2"> <li class="ltx_item" id="S4.I2.i1" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item">•</span> <div class="ltx_para" id="S4.I2.i1.p1"> <p class="ltx_p" id="S4.I2.i1.p1.1"><span class="ltx_text ltx_font_bold" id="S4.I2.i1.p1.1.1">Disease-related references</span> (e.g., ASF, PRRS, FMD, PED).</p> </div> </li> <li class="ltx_item" id="S4.I2.i2" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item">•</span> <div class="ltx_para ltx_noindent" id="S4.I2.i2.p1"> <p class="ltx_p" id="S4.I2.i2.p1.1"><span class="ltx_text ltx_font_bold" id="S4.I2.i2.p1.1.1">Vaccine-related references</span> (e.g., vitamins, disinfectants, hormones, antihelminthics).</p> </div> </li> </ul> </div> <div class="ltx_para ltx_noindent" id="S4.SS1.p4"> <p class="ltx_p" id="S4.SS1.p4.1">Each question was carefully annotated with the following metadata:</p> <ul class="ltx_itemize" id="S4.I3"> <li class="ltx_item" id="S4.I3.i1" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item">•</span> <div class="ltx_para" id="S4.I3.i1.p1"> <p class="ltx_p" id="S4.I3.i1.p1.1"><span class="ltx_text ltx_font_bold" id="S4.I3.i1.p1.1.1">Scenario:</span> The specific scenario or methodology used to generate the question (e.g., multi-turn dialogue, n-page document analysis).</p> </div> </li> <li class="ltx_item" id="S4.I3.i2" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item">•</span> <div class="ltx_para" id="S4.I3.i2.p1"> <p class="ltx_p" id="S4.I3.i2.p1.1"><span class="ltx_text ltx_font_bold" id="S4.I3.i2.p1.1.1">Question Type:</span> Categorized by complexity and structure (e.g., factoid, yes–no).</p> </div> </li> <li class="ltx_item" id="S4.I3.i3" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item">•</span> <div class="ltx_para" id="S4.I3.i3.p1"> <p class="ltx_p" id="S4.I3.i3.p1.1"><span class="ltx_text ltx_font_bold" id="S4.I3.i3.p1.1.1">Document Source:</span> The exact document and page number used as the basis for question generation.</p> </div> </li> <li class="ltx_item" id="S4.I3.i4" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item">•</span> <div class="ltx_para ltx_noindent" id="S4.I3.i4.p1"> <p class="ltx_p" id="S4.I3.i4.p1.1"><span class="ltx_text ltx_font_bold" id="S4.I3.i4.p1.1.1">Example Answer:</span> A model response demonstrating the expected level of detail and accuracy.</p> </div> </li> </ul> </div> <div class="ltx_para ltx_noindent" id="S4.SS1.p5"> <p class="ltx_p" id="S4.SS1.p5.1">This structured approach to data collection ensures that both datasets adhere to a uniform format and difficulty distribution, enabling a fair and rigorous evaluation of model performance across different question types and complexities.</p> </div> </section> <section class="ltx_subsection" id="S4.SS2"> <h3 class="ltx_title ltx_title_subsection"> <span class="ltx_tag ltx_tag_subsection">4.2 </span>Evaluation Methodology</h3> <div class="ltx_para ltx_noindent" id="S4.SS2.p1"> <p class="ltx_p" id="S4.SS2.p1.1">This section outlines the evaluation framework for assessing our AI system’s performance in Query Classification, Disease Diagnosis, and Knowledge Retrieval. We define key metrics for each task and present a comparative performance analysis.</p> </div> <section class="ltx_subsubsection" id="S4.SS2.SSS1"> <h4 class="ltx_title ltx_title_subsubsection"> <span class="ltx_tag ltx_tag_subsubsection">4.2.1 </span>Evaluation Metrics for Query Classification and Routing</h4> <div class="ltx_para ltx_noindent" id="S4.SS2.SSS1.p1"> <p class="ltx_p" id="S4.SS2.SSS1.p1.1">For the query classification task, we evaluate the system using the following metrics:</p> <ul class="ltx_itemize" id="S4.I4"> <li class="ltx_item" id="S4.I4.i1" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item">•</span> <div class="ltx_para" id="S4.I4.i1.p1"> <p class="ltx_p" id="S4.I4.i1.p1.1"><span class="ltx_text ltx_font_bold" id="S4.I4.i1.p1.1.1">Precision</span>: The proportion of correctly predicted positive cases out of all cases predicted as positive.</p> </div> </li> <li class="ltx_item" id="S4.I4.i2" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item">•</span> <div class="ltx_para" id="S4.I4.i2.p1"> <p class="ltx_p" id="S4.I4.i2.p1.1"><span class="ltx_text ltx_font_bold" id="S4.I4.i2.p1.1.1">Recall</span>: The proportion of correctly predicted positive cases out of all actual positive cases.</p> </div> </li> <li class="ltx_item" id="S4.I4.i3" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item">•</span> <div class="ltx_para" id="S4.I4.i3.p1"> <p class="ltx_p" id="S4.I4.i3.p1.1"><span class="ltx_text ltx_font_bold" id="S4.I4.i3.p1.1.1">F1-score</span>: The harmonic mean of precision and recall.</p> </div> </li> <li class="ltx_item" id="S4.I4.i4" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item">•</span> <div class="ltx_para ltx_noindent" id="S4.I4.i4.p1"> <p class="ltx_p" id="S4.I4.i4.p1.1"><span class="ltx_text ltx_font_bold" id="S4.I4.i4.p1.1.1">Accuracy</span>: The overall proportion of correctly classified cases within the dataset.</p> </div> </li> </ul> </div> </section> <section class="ltx_subsubsection" id="S4.SS2.SSS2"> <h4 class="ltx_title ltx_title_subsubsection"> <span class="ltx_tag ltx_tag_subsubsection">4.2.2 </span>Evaluation Metrics for Symptom-Based Diagnostic Queries</h4> <div class="ltx_para ltx_noindent" id="S4.SS2.SSS2.p1"> <p class="ltx_p" id="S4.SS2.SSS2.p1.1">The disease diagnosis task is evaluated using a set of metrics that capture both the accuracy of the diagnostic predictions and the efficiency of the response generation:</p> <ul class="ltx_itemize" id="S4.I5"> <li class="ltx_item" id="S4.I5.i1" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item">•</span> <div class="ltx_para" id="S4.I5.i1.p1"> <p class="ltx_p" id="S4.I5.i1.p1.1"><span class="ltx_text ltx_font_bold" id="S4.I5.i1.p1.1.1">Accuracy</span>: Models were evaluated based on diagnostic accuracy, defined as the percentage of cases in which the correct diagnosis appeared among the top two ranked predictions.</p> </div> </li> <li class="ltx_item" id="S4.I5.i2" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item">•</span> <div class="ltx_para ltx_noindent" id="S4.I5.i2.p1"> <p class="ltx_p" id="S4.I5.i2.p1.1"><span class="ltx_text ltx_font_bold" id="S4.I5.i2.p1.1.1">Execution Time</span>: The time (in seconds) taken by the model to produce its diagnostic output, reflecting computational efficiency.</p> </div> </li> </ul> </div> <div class="ltx_para ltx_noindent" id="S4.SS2.SSS2.p2"> <p class="ltx_p" id="S4.SS2.SSS2.p2.1"><span class="ltx_text ltx_font_bold" id="S4.SS2.SSS2.p2.1.1">Comparative Analysis:</span> In our pilot study, we selected three leading models: o1-mini, GPT-4o, and Gemini-1.5-Pro-002—for disease diagnosis. Their selection is supported by:</p> <ul class="ltx_itemize" id="S4.I6"> <li class="ltx_item" id="S4.I6.i1" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item">•</span> <div class="ltx_para" id="S4.I6.i1.p1"> <p class="ltx_p" id="S4.I6.i1.p1.1"><span class="ltx_text ltx_font_bold" id="S4.I6.i1.p1.1.1">o1-mini:</span> Exhibits efficient language understanding and rapid context retention, making it suitable for streamlined diagnostic applications <cite class="ltx_cite ltx_citemacro_cite">[<a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#bib.bib21" title="">21</a>]</cite>.</p> </div> </li> <li class="ltx_item" id="S4.I6.i2" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item">•</span> <div class="ltx_para" id="S4.I6.i2.p1"> <p class="ltx_p" id="S4.I6.i2.p1.1"><span class="ltx_text ltx_font_bold" id="S4.I6.i2.p1.1.1">GPT-4o:</span> Provides advanced integration of diverse clinical data, with its capabilities outlined in recent evaluations <cite class="ltx_cite ltx_citemacro_cite">[<a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#bib.bib22" title="">22</a>]</cite>.</p> </div> </li> <li class="ltx_item" id="S4.I6.i3" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item">•</span> <div class="ltx_para ltx_noindent" id="S4.I6.i3.p1"> <p class="ltx_p" id="S4.I6.i3.p1.1"><span class="ltx_text ltx_font_bold" id="S4.I6.i3.p1.1.1">Gemini-1.5-Pro-002:</span> Employs a multimodal approach tailored for medical-specific inputs, showing promising diagnostic precision in early assessments <cite class="ltx_cite ltx_citemacro_cite">[<a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#bib.bib23" title="">23</a>]</cite>.</p> </div> </li> </ul> </div> </section> <section class="ltx_subsubsection" id="S4.SS2.SSS3"> <h4 class="ltx_title ltx_title_subsubsection"> <span class="ltx_tag ltx_tag_subsubsection">4.2.3 </span>Evaluation Metrics for Knowledge Retrieval Queries</h4> <div class="ltx_para ltx_noindent" id="S4.SS2.SSS3.p1"> <p class="ltx_p" id="S4.SS2.SSS3.p1.1">For the knowledge retrieval task, responses are evaluated using a tailored scoring system (ranging from 0 to 5) based on the following dimensions:</p> <ul class="ltx_itemize" id="S4.I7"> <li class="ltx_item" id="S4.I7.i1" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item">•</span> <div class="ltx_para" id="S4.I7.i1.p1"> <p class="ltx_p" id="S4.I7.i1.p1.1"><span class="ltx_text ltx_font_bold" id="S4.I7.i1.p1.1.1">Expansiveness</span>: The extent to which the response thoroughly covers the topic.</p> </div> </li> <li class="ltx_item" id="S4.I7.i2" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item">•</span> <div class="ltx_para" id="S4.I7.i2.p1"> <p class="ltx_p" id="S4.I7.i2.p1.1"><span class="ltx_text ltx_font_bold" id="S4.I7.i2.p1.1.1">Coherence</span>: The logical organization and readability of the response.</p> </div> </li> <li class="ltx_item" id="S4.I7.i3" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item">•</span> <div class="ltx_para" id="S4.I7.i3.p1"> <p class="ltx_p" id="S4.I7.i3.p1.1"><span class="ltx_text ltx_font_bold" id="S4.I7.i3.p1.1.1">Correctness</span>: The assessment of whether the response is overall accurate and well-supported by evidence.</p> </div> </li> <li class="ltx_item" id="S4.I7.i4" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item">•</span> <div class="ltx_para" id="S4.I7.i4.p1"> <p class="ltx_p" id="S4.I7.i4.p1.1"><span class="ltx_text ltx_font_bold" id="S4.I7.i4.p1.1.1">Relevance</span>: The degree to which the response directly addresses the query.</p> </div> </li> <li class="ltx_item" id="S4.I7.i5" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item">•</span> <div class="ltx_para ltx_noindent" id="S4.I7.i5.p1"> <p class="ltx_p" id="S4.I7.i5.p1.1"><span class="ltx_text ltx_font_bold" id="S4.I7.i5.p1.1.1">Accuracy</span>: The correctness of the response based on the question type, ensuring it is free from misinformation.</p> </div> </li> </ul> </div> <div class="ltx_para ltx_noindent" id="S4.SS2.SSS3.p2"> <p class="ltx_p" id="S4.SS2.SSS3.p2.1"><span class="ltx_text ltx_font_bold" id="S4.SS2.SSS3.p2.1.1">Comparative Analysis:</span> To evaluate improvements in our knowledge retrieval system, we compared it against the approach in <cite class="ltx_cite ltx_citemacro_cite">[<a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#bib.bib24" title="">24</a>]</cite>, a global platform for generalized retrieval tasks. Using validation and test datasets, we analyzed average performance scores and conducted paired t-tests on matched samples with a 0.05 significance threshold. To enhance reliability, we applied bootstrapping, sampling 80% of matched data points repeatedly and averaging the t-statistics and p-values.</p> </div> </section> </section> </section> <section class="ltx_section" id="S5"> <h2 class="ltx_title ltx_title_section"> <span class="ltx_tag ltx_tag_section">5 </span>Results</h2> <section class="ltx_subsection" id="S5.SS1"> <h3 class="ltx_title ltx_title_subsection"> <span class="ltx_tag ltx_tag_subsection">5.1 </span>Query Classification Results</h3> <div class="ltx_para ltx_noindent" id="S5.SS1.p1"> <p class="ltx_p" id="S5.SS1.p1.1">The table <a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#S5.T2" title="Table 2 ‣ 5.1 Query Classification Results ‣ 5 Results ‣ When Pigs Get Sick: Multi-Agent AI for Swine Disease Detection"><span class="ltx_text ltx_ref_tag">2</span></a> presents the classification performance of the system across four query types: General, Retrieval (Knowledge Retrieval), Diagnosis (Symptom-Based Diagnostic), and TBC. The system achieves an overall accuracy of <span class="ltx_text ltx_font_bold" id="S5.SS1.p1.1.1">95.23%</span>, demonstrating strong performance in correctly categorizing user queries.</p> </div> <figure class="ltx_table" id="S5.T2"> <table class="ltx_tabular ltx_centering ltx_align_middle" id="S5.T2.3"> <tbody class="ltx_tbody"> <tr class="ltx_tr" id="S5.T2.3.4.1"> <td class="ltx_td ltx_align_center ltx_border_l ltx_border_r ltx_border_t" id="S5.T2.3.4.1.1" rowspan="2" style="padding-top:1.5pt;padding-bottom:1.5pt;"><span class="ltx_text ltx_font_bold" id="S5.T2.3.4.1.1.1">Ground Truth</span></td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" colspan="4" id="S5.T2.3.4.1.2" style="padding-top:1.5pt;padding-bottom:1.5pt;"><span class="ltx_text ltx_font_bold" id="S5.T2.3.4.1.2.1">Prediction</span></td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T2.3.4.1.3" rowspan="2" style="padding-top:1.5pt;padding-bottom:1.5pt;"><span class="ltx_text ltx_font_bold" id="S5.T2.3.4.1.3.1">Precision</span></td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T2.3.4.1.4" rowspan="2" style="padding-top:1.5pt;padding-bottom:1.5pt;"><span class="ltx_text ltx_font_bold" id="S5.T2.3.4.1.4.1">Recall</span></td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T2.3.4.1.5" rowspan="2" style="padding-top:1.5pt;padding-bottom:1.5pt;"><span class="ltx_text ltx_font_bold" id="S5.T2.3.4.1.5.1">F1-score</span></td> </tr> <tr class="ltx_tr" id="S5.T2.3.5.2"> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T2.3.5.2.1" style="padding-top:1.5pt;padding-bottom:1.5pt;"><span class="ltx_text ltx_font_bold" id="S5.T2.3.5.2.1.1">General</span></td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T2.3.5.2.2" style="padding-top:1.5pt;padding-bottom:1.5pt;"><span class="ltx_text ltx_font_bold" id="S5.T2.3.5.2.2.1">Retrieval</span></td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T2.3.5.2.3" style="padding-top:1.5pt;padding-bottom:1.5pt;"><span class="ltx_text ltx_font_bold" id="S5.T2.3.5.2.3.1">Diagnosis</span></td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T2.3.5.2.4" style="padding-top:1.5pt;padding-bottom:1.5pt;"><span class="ltx_text ltx_font_bold" id="S5.T2.3.5.2.4.1">TBC</span></td> </tr> <tr class="ltx_tr" id="S5.T2.1.1"> <td class="ltx_td ltx_align_center ltx_border_l ltx_border_r ltx_border_t" id="S5.T2.1.1.2" style="padding-top:1.5pt;padding-bottom:1.5pt;"><span class="ltx_text ltx_font_bold" id="S5.T2.1.1.2.1">General</span></td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T2.1.1.3" style="padding-top:1.5pt;padding-bottom:1.5pt;"><span class="ltx_text ltx_font_bold" id="S5.T2.1.1.3.1">24</span></td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T2.1.1.4" style="padding-top:1.5pt;padding-bottom:1.5pt;">0</td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T2.1.1.5" style="padding-top:1.5pt;padding-bottom:1.5pt;">0</td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T2.1.1.6" style="padding-top:1.5pt;padding-bottom:1.5pt;">0</td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T2.1.1.7" style="padding-top:1.5pt;padding-bottom:1.5pt;">0.750</td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T2.1.1.1" style="padding-top:1.5pt;padding-bottom:1.5pt;"><span class="ltx_text ltx_font_bold" id="S5.T2.1.1.1.1">1.000<sup class="ltx_sup" id="S5.T2.1.1.1.1.1"><span class="ltx_text ltx_font_medium" id="S5.T2.1.1.1.1.1.1">†</span></sup></span></td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T2.1.1.8" style="padding-top:1.5pt;padding-bottom:1.5pt;">0.857</td> </tr> <tr class="ltx_tr" id="S5.T2.3.3"> <td class="ltx_td ltx_align_center ltx_border_l ltx_border_r ltx_border_t" id="S5.T2.3.3.3" style="padding-top:1.5pt;padding-bottom:1.5pt;"><span class="ltx_text ltx_font_bold" id="S5.T2.3.3.3.1">Retrieval</span></td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T2.3.3.4" style="padding-top:1.5pt;padding-bottom:1.5pt;">6</td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T2.3.3.5" style="padding-top:1.5pt;padding-bottom:1.5pt;"><span class="ltx_text ltx_font_bold" id="S5.T2.3.3.5.1">301</span></td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T2.3.3.6" style="padding-top:1.5pt;padding-bottom:1.5pt;">5</td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T2.3.3.7" style="padding-top:1.5pt;padding-bottom:1.5pt;">6</td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T2.2.2.1" style="padding-top:1.5pt;padding-bottom:1.5pt;"><span class="ltx_text ltx_font_bold" id="S5.T2.2.2.1.1">0.993<sup class="ltx_sup" id="S5.T2.2.2.1.1.1"><span class="ltx_text ltx_font_medium" id="S5.T2.2.2.1.1.1.1">†</span></sup></span></td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T2.3.3.8" style="padding-top:1.5pt;padding-bottom:1.5pt;">0.947</td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T2.3.3.2" style="padding-top:1.5pt;padding-bottom:1.5pt;"><span class="ltx_text ltx_font_bold" id="S5.T2.3.3.2.1">0.969<sup class="ltx_sup" id="S5.T2.3.3.2.1.1"><span class="ltx_text ltx_font_medium" id="S5.T2.3.3.2.1.1.1">†</span></sup></span></td> </tr> <tr class="ltx_tr" id="S5.T2.3.6.3"> <td class="ltx_td ltx_align_center ltx_border_l ltx_border_r ltx_border_t" id="S5.T2.3.6.3.1" style="padding-top:1.5pt;padding-bottom:1.5pt;"><span class="ltx_text ltx_font_bold" id="S5.T2.3.6.3.1.1">Diagnosis</span></td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T2.3.6.3.2" style="padding-top:1.5pt;padding-bottom:1.5pt;">0</td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T2.3.6.3.3" style="padding-top:1.5pt;padding-bottom:1.5pt;">2</td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T2.3.6.3.4" style="padding-top:1.5pt;padding-bottom:1.5pt;"><span class="ltx_text ltx_font_bold" id="S5.T2.3.6.3.4.1">82</span></td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T2.3.6.3.5" style="padding-top:1.5pt;padding-bottom:1.5pt;">1</td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T2.3.6.3.6" style="padding-top:1.5pt;padding-bottom:1.5pt;">0.943</td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T2.3.6.3.7" style="padding-top:1.5pt;padding-bottom:1.5pt;">0.965</td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T2.3.6.3.8" style="padding-top:1.5pt;padding-bottom:1.5pt;">0.953</td> </tr> <tr class="ltx_tr" id="S5.T2.3.7.4"> <td class="ltx_td ltx_align_center ltx_border_l ltx_border_r ltx_border_t" id="S5.T2.3.7.4.1" style="padding-top:1.5pt;padding-bottom:1.5pt;"><span class="ltx_text ltx_font_bold" id="S5.T2.3.7.4.1.1">TBC</span></td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T2.3.7.4.2" style="padding-top:1.5pt;padding-bottom:1.5pt;">2</td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T2.3.7.4.3" style="padding-top:1.5pt;padding-bottom:1.5pt;">0</td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T2.3.7.4.4" style="padding-top:1.5pt;padding-bottom:1.5pt;">0</td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T2.3.7.4.5" style="padding-top:1.5pt;padding-bottom:1.5pt;"><span class="ltx_text ltx_font_bold" id="S5.T2.3.7.4.5.1">32</span></td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T2.3.7.4.6" style="padding-top:1.5pt;padding-bottom:1.5pt;">0.821</td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T2.3.7.4.7" style="padding-top:1.5pt;padding-bottom:1.5pt;">0.941</td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T2.3.7.4.8" style="padding-top:1.5pt;padding-bottom:1.5pt;">0.877</td> </tr> <tr class="ltx_tr" id="S5.T2.3.8.5"> <td class="ltx_td ltx_border_b ltx_border_l ltx_border_r ltx_border_t" colspan="5" id="S5.T2.3.8.5.1" style="padding-top:1.5pt;padding-bottom:1.5pt;"></td> <td class="ltx_td ltx_align_center ltx_border_b ltx_border_r ltx_border_t" colspan="3" id="S5.T2.3.8.5.2" style="background-color:#E6E6E6;padding-top:1.5pt;padding-bottom:1.5pt;"><span class="ltx_text" id="S5.T2.3.8.5.2.1" style="background-color:#E6E6E6;"><span class="ltx_text ltx_font_bold" id="S5.T2.3.8.5.2.1.1">Accuracy : 95.23%</span></span></td> </tr> </tbody> </table> <figcaption class="ltx_caption ltx_centering"><span class="ltx_tag ltx_tag_table">Table 2: </span>Classification performance on the test set. The best values are marked with <sup class="ltx_sup" id="S5.T2.7.1">†</sup>.</figcaption> </figure> <div class="ltx_para ltx_noindent" id="S5.SS1.p2"> <p class="ltx_p" id="S5.SS1.p2.1">The classification system performs well across different query types. General queries maintain a perfect recall of 1.000, ensuring all such queries are correctly identified, though the precision is slightly lower at 0.750 due to some misclassifications. This results in an F1-score of 0.857, indicating a good balance between precision and recall, although some ambiguous queries were mistakenly classified as General instead of TBC.</p> </div> <div class="ltx_para ltx_noindent" id="S5.SS1.p3"> <p class="ltx_p" id="S5.SS1.p3.1">Retrieval queries exhibit exceptional performance, with a high precision of 0.993 and recall of 0.947, meaning the system efficiently retrieves factual information. However, a few cases were misclassified, particularly as symptom-based diagnostic queries.</p> </div> <div class="ltx_para ltx_noindent" id="S5.SS1.p4"> <p class="ltx_p" id="S5.SS1.p4.1">Diagnosis queries achieve a precision of 0.943 and a recall of 0.965, signifying that most symptom-based queries are correctly identified. Minor misclassifications occur, particularly into Retrieval queries, when symptom descriptions resemble knowledge-seeking questions. Despite this, the F1-score of 0.953 reflects strong diagnostic capability.</p> </div> <div class="ltx_para ltx_noindent" id="S5.SS1.p5"> <p class="ltx_p" id="S5.SS1.p5.1">Lastly, TBC queries are well-handled, with a precision of 0.821 and a recall of 0.941, demonstrating the system’s effectiveness in managing ambiguous and incomplete queries. However, some TBC queries may be mistakenly classified as General or Diagnosis due to insufficient context in user input.</p> </div> </section> <section class="ltx_subsection" id="S5.SS2"> <h3 class="ltx_title ltx_title_subsection"> <span class="ltx_tag ltx_tag_subsection">5.2 </span>Symptom-Based Diagnostic Results</h3> <div class="ltx_para ltx_noindent" id="S5.SS2.p1"> <p class="ltx_p" id="S5.SS2.p1.1">As shown in Table <a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#S5.T3" title="Table 3 ‣ 5.2 Symptom-Based Diagnostic Results ‣ 5 Results ‣ When Pigs Get Sick: Multi-Agent AI for Swine Disease Detection"><span class="ltx_text ltx_ref_tag">3</span></a>, although all models achieve similar test accuracy, execution times differ significantly. Notably, <span class="ltx_text ltx_font_bold" id="S5.SS2.p1.1.1">GPT-4o</span> achieves the highest test accuracy at <span class="ltx_text ltx_font_bold" id="S5.SS2.p1.1.2">90.63%</span> while also being the fastest, with an execution time of <span class="ltx_text ltx_font_bold" id="S5.SS2.p1.1.3">18.78</span> seconds. Its validation accuracy of 88.46% remains close, demonstrating consistency and robustness across datasets. In contrast, Gemini-1.5-Pro-002 achieves the highest validation accuracy at 94.23% but drops to 87.50% on the test set, suggesting potential generalization issues. Meanwhile, o1-mini-mini lags in both accuracy and speed, with the highest latency at 29.38 seconds, which limits its practical usability. Additionally, ASF and PED yield the most reliable diagnoses, whereas PRRS and FMD exhibit inconsistencies. A deeper analysis in Section <a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#S6.SS2" title="6.2 Misclassification of Diseases ‣ 6 Discussion ‣ When Pigs Get Sick: Multi-Agent AI for Swine Disease Detection"><span class="ltx_text ltx_ref_tag">6.2</span></a> will explore potential reasons for these variations.</p> </div> <figure class="ltx_table" id="S5.T3"> <table class="ltx_tabular ltx_centering ltx_guessed_headers ltx_align_middle" id="S5.T3.4"> <tbody class="ltx_tbody"> <tr class="ltx_tr" id="S5.T3.4.5.1"> <th class="ltx_td ltx_align_left ltx_th ltx_th_row ltx_border_l ltx_border_r ltx_border_t" id="S5.T3.4.5.1.1" rowspan="2" style="padding-top:1.5pt;padding-bottom:1.5pt;"><span class="ltx_text ltx_font_bold" id="S5.T3.4.5.1.1.1">Model</span></th> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T3.4.5.1.2" rowspan="2" style="padding-top:1.5pt;padding-bottom:1.5pt;"><span class="ltx_text ltx_font_bold" id="S5.T3.4.5.1.2.1">Disease</span></td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" colspan="2" id="S5.T3.4.5.1.3" style="padding-top:1.5pt;padding-bottom:1.5pt;"><span class="ltx_text ltx_font_bold" id="S5.T3.4.5.1.3.1">Predicted / Actual</span></td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" colspan="2" id="S5.T3.4.5.1.4" style="padding-top:1.5pt;padding-bottom:1.5pt;"><span class="ltx_text ltx_font_bold" id="S5.T3.4.5.1.4.1">Execution Time (s)</span></td> </tr> <tr class="ltx_tr" id="S5.T3.4.6.2"> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T3.4.6.2.1" style="padding-top:1.5pt;padding-bottom:1.5pt;"><span class="ltx_text ltx_font_bold" id="S5.T3.4.6.2.1.1">Validation</span></td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T3.4.6.2.2" style="padding-top:1.5pt;padding-bottom:1.5pt;"><span class="ltx_text ltx_font_bold" id="S5.T3.4.6.2.2.1">Test</span></td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T3.4.6.2.3" style="padding-top:1.5pt;padding-bottom:1.5pt;"><span class="ltx_text ltx_font_bold" id="S5.T3.4.6.2.3.1">Validation</span></td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T3.4.6.2.4" style="padding-top:1.5pt;padding-bottom:1.5pt;"><span class="ltx_text ltx_font_bold" id="S5.T3.4.6.2.4.1">Test</span></td> </tr> <tr class="ltx_tr" id="S5.T3.4.7.3"> <th class="ltx_td ltx_align_left ltx_th ltx_th_row ltx_border_l ltx_border_r ltx_border_t" id="S5.T3.4.7.3.1" rowspan="5" style="padding-top:1.5pt;padding-bottom:1.5pt;"><span class="ltx_text ltx_font_bold" id="S5.T3.4.7.3.1.1">GPT-4o</span></th> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T3.4.7.3.2" style="padding-top:1.5pt;padding-bottom:1.5pt;">ASF</td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T3.4.7.3.3" style="padding-top:1.5pt;padding-bottom:1.5pt;">(7/7)</td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T3.4.7.3.4" style="padding-top:1.5pt;padding-bottom:1.5pt;">(5/5)</td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T3.4.7.3.5" style="padding-top:1.5pt;padding-bottom:1.5pt;">20.09</td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T3.4.7.3.6" style="padding-top:1.5pt;padding-bottom:1.5pt;">17.83</td> </tr> <tr class="ltx_tr" id="S5.T3.4.8.4"> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T3.4.8.4.1" style="padding-top:1.5pt;padding-bottom:1.5pt;">PRRS</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T3.4.8.4.2" style="padding-top:1.5pt;padding-bottom:1.5pt;">(4/5)</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T3.4.8.4.3" style="padding-top:1.5pt;padding-bottom:1.5pt;">(11/14)</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T3.4.8.4.4" style="padding-top:1.5pt;padding-bottom:1.5pt;">19.20</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T3.4.8.4.5" style="padding-top:1.5pt;padding-bottom:1.5pt;">18.82</td> </tr> <tr class="ltx_tr" id="S5.T3.4.9.5"> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T3.4.9.5.1" style="padding-top:1.5pt;padding-bottom:1.5pt;">PED</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T3.4.9.5.2" style="padding-top:1.5pt;padding-bottom:1.5pt;">(4/4)</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T3.4.9.5.3" style="padding-top:1.5pt;padding-bottom:1.5pt;">(5/5)</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T3.4.9.5.4" style="padding-top:1.5pt;padding-bottom:1.5pt;">19.02</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T3.4.9.5.5" style="padding-top:1.5pt;padding-bottom:1.5pt;">20.17</td> </tr> <tr class="ltx_tr" id="S5.T3.4.10.6"> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T3.4.10.6.1" style="padding-top:1.5pt;padding-bottom:1.5pt;">FMD</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T3.4.10.6.2" style="padding-top:1.5pt;padding-bottom:1.5pt;">(3/6)</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T3.4.10.6.3" style="padding-top:1.5pt;padding-bottom:1.5pt;">(7/7)</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T3.4.10.6.4" style="padding-top:1.5pt;padding-bottom:1.5pt;">19.94</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T3.4.10.6.5" style="padding-top:1.5pt;padding-bottom:1.5pt;">18.75</td> </tr> <tr class="ltx_tr" id="S5.T3.4.11.7"> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T3.4.11.7.1" style="padding-top:1.5pt;padding-bottom:1.5pt;">OOD</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T3.4.11.7.2" style="padding-top:1.5pt;padding-bottom:1.5pt;">(28/30)</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T3.4.11.7.3" style="padding-top:1.5pt;padding-bottom:1.5pt;">(1/1)</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T3.4.11.7.4" style="padding-top:1.5pt;padding-bottom:1.5pt;">19.53</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T3.4.11.7.5" style="padding-top:1.5pt;padding-bottom:1.5pt;">18.31</td> </tr> <tr class="ltx_tr" id="S5.T3.3.3" style="background-color:#E6E6E6;"> <th class="ltx_td ltx_align_left ltx_th ltx_th_row ltx_border_l ltx_border_r ltx_border_t" id="S5.T3.3.3.4" style="padding-top:1.5pt;padding-bottom:1.5pt;"><span class="ltx_text ltx_font_bold" id="S5.T3.3.3.4.1" style="background-color:#E6E6E6;">Accuracy (%)</span></th> <td class="ltx_td ltx_border_r ltx_border_t" id="S5.T3.3.3.5" style="padding-top:1.5pt;padding-bottom:1.5pt;"></td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T3.3.3.6" style="padding-top:1.5pt;padding-bottom:1.5pt;"><span class="ltx_text" id="S5.T3.3.3.6.1" style="background-color:#E6E6E6;">88.46</span></td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T3.1.1.1" style="padding-top:1.5pt;padding-bottom:1.5pt;"><span class="ltx_text ltx_font_bold" id="S5.T3.1.1.1.1" style="background-color:#E6E6E6;">90.63<sup class="ltx_sup" id="S5.T3.1.1.1.1.1"><span class="ltx_text ltx_font_medium" id="S5.T3.1.1.1.1.1.1">†</span></sup></span></td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T3.2.2.2" style="padding-top:1.5pt;padding-bottom:1.5pt;"><span class="ltx_text ltx_font_bold" id="S5.T3.2.2.2.1" style="background-color:#E6E6E6;">19.55<sup class="ltx_sup" id="S5.T3.2.2.2.1.1"><span class="ltx_text ltx_font_medium" id="S5.T3.2.2.2.1.1.1">†</span></sup></span></td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T3.3.3.3" style="padding-top:1.5pt;padding-bottom:1.5pt;"><span class="ltx_text ltx_font_bold" id="S5.T3.3.3.3.1" style="background-color:#E6E6E6;">18.78<sup class="ltx_sup" id="S5.T3.3.3.3.1.1"><span class="ltx_text ltx_font_medium" id="S5.T3.3.3.3.1.1.1">†</span></sup></span></td> </tr> <tr class="ltx_tr" id="S5.T3.4.12.8"> <th class="ltx_td ltx_align_left ltx_th ltx_th_row ltx_border_l ltx_border_r ltx_border_t" id="S5.T3.4.12.8.1" rowspan="5" style="padding-top:1.5pt;padding-bottom:1.5pt;"><span class="ltx_text ltx_font_bold" id="S5.T3.4.12.8.1.1">o1-mini-mini</span></th> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T3.4.12.8.2" style="padding-top:1.5pt;padding-bottom:1.5pt;">ASF</td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T3.4.12.8.3" style="padding-top:1.5pt;padding-bottom:1.5pt;">(7/7)</td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T3.4.12.8.4" style="padding-top:1.5pt;padding-bottom:1.5pt;">(5/5)</td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T3.4.12.8.5" style="padding-top:1.5pt;padding-bottom:1.5pt;">27.69</td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T3.4.12.8.6" style="padding-top:1.5pt;padding-bottom:1.5pt;">28.55</td> </tr> <tr class="ltx_tr" id="S5.T3.4.13.9"> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T3.4.13.9.1" style="padding-top:1.5pt;padding-bottom:1.5pt;">PRRS</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T3.4.13.9.2" style="padding-top:1.5pt;padding-bottom:1.5pt;">(5/5)</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T3.4.13.9.3" style="padding-top:1.5pt;padding-bottom:1.5pt;">(10/14)</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T3.4.13.9.4" style="padding-top:1.5pt;padding-bottom:1.5pt;">31.03</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T3.4.13.9.5" style="padding-top:1.5pt;padding-bottom:1.5pt;">29.19</td> </tr> <tr class="ltx_tr" id="S5.T3.4.14.10"> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T3.4.14.10.1" style="padding-top:1.5pt;padding-bottom:1.5pt;">PED</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T3.4.14.10.2" style="padding-top:1.5pt;padding-bottom:1.5pt;">(4/4)</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T3.4.14.10.3" style="padding-top:1.5pt;padding-bottom:1.5pt;">(5/5)</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T3.4.14.10.4" style="padding-top:1.5pt;padding-bottom:1.5pt;">31.38</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T3.4.14.10.5" style="padding-top:1.5pt;padding-bottom:1.5pt;">31.24</td> </tr> <tr class="ltx_tr" id="S5.T3.4.15.11"> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T3.4.15.11.1" style="padding-top:1.5pt;padding-bottom:1.5pt;">FMD</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T3.4.15.11.2" style="padding-top:1.5pt;padding-bottom:1.5pt;">(3/6)</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T3.4.15.11.3" style="padding-top:1.5pt;padding-bottom:1.5pt;">(6/7)</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T3.4.15.11.4" style="padding-top:1.5pt;padding-bottom:1.5pt;">27.89</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T3.4.15.11.5" style="padding-top:1.5pt;padding-bottom:1.5pt;">27.57</td> </tr> <tr class="ltx_tr" id="S5.T3.4.16.12"> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T3.4.16.12.1" style="padding-top:1.5pt;padding-bottom:1.5pt;">OOD</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T3.4.16.12.2" style="padding-top:1.5pt;padding-bottom:1.5pt;">(24/30)</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T3.4.16.12.3" style="padding-top:1.5pt;padding-bottom:1.5pt;">(1/1)</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T3.4.16.12.4" style="padding-top:1.5pt;padding-bottom:1.5pt;">28.91</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T3.4.16.12.5" style="padding-top:1.5pt;padding-bottom:1.5pt;">26.57</td> </tr> <tr class="ltx_tr" id="S5.T3.4.17.13" style="background-color:#E6E6E6;"> <th class="ltx_td ltx_align_left ltx_th ltx_th_row ltx_border_l ltx_border_r ltx_border_t" id="S5.T3.4.17.13.1" style="padding-top:1.5pt;padding-bottom:1.5pt;"><span class="ltx_text ltx_font_bold" id="S5.T3.4.17.13.1.1" style="background-color:#E6E6E6;">Accuracy (%)</span></th> <td class="ltx_td ltx_border_r ltx_border_t" id="S5.T3.4.17.13.2" style="padding-top:1.5pt;padding-bottom:1.5pt;"></td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T3.4.17.13.3" style="padding-top:1.5pt;padding-bottom:1.5pt;"><span class="ltx_text" id="S5.T3.4.17.13.3.1" style="background-color:#E6E6E6;">82.69</span></td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T3.4.17.13.4" style="padding-top:1.5pt;padding-bottom:1.5pt;"><span class="ltx_text" id="S5.T3.4.17.13.4.1" style="background-color:#E6E6E6;">84.37</span></td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T3.4.17.13.5" style="padding-top:1.5pt;padding-bottom:1.5pt;"><span class="ltx_text" id="S5.T3.4.17.13.5.1" style="background-color:#E6E6E6;">29.38</span></td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T3.4.17.13.6" style="padding-top:1.5pt;padding-bottom:1.5pt;"><span class="ltx_text" id="S5.T3.4.17.13.6.1" style="background-color:#E6E6E6;">28.63</span></td> </tr> <tr class="ltx_tr" id="S5.T3.4.18.14"> <th class="ltx_td ltx_align_left ltx_th ltx_th_row ltx_border_l ltx_border_r ltx_border_t" id="S5.T3.4.18.14.1" rowspan="5" style="padding-top:1.5pt;padding-bottom:1.5pt;"><span class="ltx_text ltx_font_bold" id="S5.T3.4.18.14.1.1">Gemini-1.5</span></th> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T3.4.18.14.2" style="padding-top:1.5pt;padding-bottom:1.5pt;">ASF</td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T3.4.18.14.3" style="padding-top:1.5pt;padding-bottom:1.5pt;">(7/7)</td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T3.4.18.14.4" style="padding-top:1.5pt;padding-bottom:1.5pt;">(5/5)</td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T3.4.18.14.5" style="padding-top:1.5pt;padding-bottom:1.5pt;">22.19</td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T3.4.18.14.6" style="padding-top:1.5pt;padding-bottom:1.5pt;">21.87</td> </tr> <tr class="ltx_tr" id="S5.T3.4.19.15"> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T3.4.19.15.1" style="padding-top:1.5pt;padding-bottom:1.5pt;">PRRS</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T3.4.19.15.2" style="padding-top:1.5pt;padding-bottom:1.5pt;">(5/5)</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T3.4.19.15.3" style="padding-top:1.5pt;padding-bottom:1.5pt;">(11/14)</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T3.4.19.15.4" style="padding-top:1.5pt;padding-bottom:1.5pt;">22.42</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T3.4.19.15.5" style="padding-top:1.5pt;padding-bottom:1.5pt;">24.62</td> </tr> <tr class="ltx_tr" id="S5.T3.4.20.16"> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T3.4.20.16.1" style="padding-top:1.5pt;padding-bottom:1.5pt;">PED</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T3.4.20.16.2" style="padding-top:1.5pt;padding-bottom:1.5pt;">(4/4)</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T3.4.20.16.3" style="padding-top:1.5pt;padding-bottom:1.5pt;">(5/5)</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T3.4.20.16.4" style="padding-top:1.5pt;padding-bottom:1.5pt;">21.58</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T3.4.20.16.5" style="padding-top:1.5pt;padding-bottom:1.5pt;">26.31</td> </tr> <tr class="ltx_tr" id="S5.T3.4.21.17"> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T3.4.21.17.1" style="padding-top:1.5pt;padding-bottom:1.5pt;">FMD</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T3.4.21.17.2" style="padding-top:1.5pt;padding-bottom:1.5pt;">(3/6)</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T3.4.21.17.3" style="padding-top:1.5pt;padding-bottom:1.5pt;">(6/7)</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T3.4.21.17.4" style="padding-top:1.5pt;padding-bottom:1.5pt;">23.52</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T3.4.21.17.5" style="padding-top:1.5pt;padding-bottom:1.5pt;">20.62</td> </tr> <tr class="ltx_tr" id="S5.T3.4.22.18"> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T3.4.22.18.1" style="padding-top:1.5pt;padding-bottom:1.5pt;">OOD</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T3.4.22.18.2" style="padding-top:1.5pt;padding-bottom:1.5pt;">(30/30)</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T3.4.22.18.3" style="padding-top:1.5pt;padding-bottom:1.5pt;">(1/1)</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T3.4.22.18.4" style="padding-top:1.5pt;padding-bottom:1.5pt;">21.82</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T3.4.22.18.5" style="padding-top:1.5pt;padding-bottom:1.5pt;">20.30</td> </tr> <tr class="ltx_tr" id="S5.T3.4.4" style="background-color:#E6E6E6;"> <th class="ltx_td ltx_align_left ltx_th ltx_th_row ltx_border_b ltx_border_l ltx_border_r ltx_border_t" id="S5.T3.4.4.2" style="padding-top:1.5pt;padding-bottom:1.5pt;"><span class="ltx_text ltx_font_bold" id="S5.T3.4.4.2.1" style="background-color:#E6E6E6;">Accuracy (%)</span></th> <td class="ltx_td ltx_border_b ltx_border_r ltx_border_t" id="S5.T3.4.4.3" style="padding-top:1.5pt;padding-bottom:1.5pt;"></td> <td class="ltx_td ltx_align_center ltx_border_b ltx_border_r ltx_border_t" id="S5.T3.4.4.1" style="padding-top:1.5pt;padding-bottom:1.5pt;"><span class="ltx_text ltx_font_bold" id="S5.T3.4.4.1.1" style="background-color:#E6E6E6;">94.23<sup class="ltx_sup" id="S5.T3.4.4.1.1.1"><span class="ltx_text ltx_font_medium" id="S5.T3.4.4.1.1.1.1">†</span></sup></span></td> <td class="ltx_td ltx_align_center ltx_border_b ltx_border_r ltx_border_t" id="S5.T3.4.4.4" style="padding-top:1.5pt;padding-bottom:1.5pt;"><span class="ltx_text" id="S5.T3.4.4.4.1" style="background-color:#E6E6E6;">87.50</span></td> <td class="ltx_td ltx_align_center ltx_border_b ltx_border_r ltx_border_t" id="S5.T3.4.4.5" style="padding-top:1.5pt;padding-bottom:1.5pt;"><span class="ltx_text" id="S5.T3.4.4.5.1" style="background-color:#E6E6E6;">22.30</span></td> <td class="ltx_td ltx_align_center ltx_border_b ltx_border_r ltx_border_t" id="S5.T3.4.4.6" style="padding-top:1.5pt;padding-bottom:1.5pt;"><span class="ltx_text" id="S5.T3.4.4.6.1" style="background-color:#E6E6E6;">22.74</span></td> </tr> </tbody> </table> <figcaption class="ltx_caption ltx_centering"><span class="ltx_tag ltx_tag_table">Table 3: </span>Performance comparison of different models on disease diagnosis tasks.</figcaption> </figure> <figure class="ltx_figure" id="S5.F2"> <div class="ltx_flex_figure"> <div class="ltx_flex_cell ltx_flex_size_2"> <figure class="ltx_figure ltx_figure_panel ltx_align_center" id="S5.F2.sf1"><img alt="Refer to caption" class="ltx_graphics ltx_img_landscape" height="165" id="S5.F2.sf1.g1" src="extracted/6293469/fig/conf_dis.png" width="269"/> <figcaption class="ltx_caption"><span class="ltx_tag ltx_tag_figure">(a) </span>Confidence Distribution for Correct Predictions</figcaption> </figure> </div> <div class="ltx_flex_cell ltx_flex_size_2"> <figure class="ltx_figure ltx_figure_panel ltx_align_center" id="S5.F2.sf2"><img alt="Refer to caption" class="ltx_graphics ltx_img_landscape" height="161" id="S5.F2.sf2.g1" src="extracted/6293469/fig/score_dis.png" width="269"/> <figcaption class="ltx_caption"><span class="ltx_tag ltx_tag_figure">(b) </span>Score Distributions across Key Performance Metrics</figcaption> </figure> </div> </div> <figcaption class="ltx_caption ltx_centering"><span class="ltx_tag ltx_tag_figure">Figure 2: </span>Subfigure (a) illustrates the distribution of model confidence scores specifically for correct predictions, offering insights into the relative certainty of outputs across different models. Subfigure (b) presents the performance ratings distribution across key evaluation metrics, highlighting the model’s consistent and strong performance in accuracy, relevance, correctness, coherence, and expansiveness.</figcaption> </figure> <div class="ltx_para ltx_noindent" id="S5.SS2.p2"> <p class="ltx_p" id="S5.SS2.p2.1">To further analyze model behavior, Figure <a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#S5.F2.sf1" title="In Figure 2 ‣ 5.2 Symptom-Based Diagnostic Results ‣ 5 Results ‣ When Pigs Get Sick: Multi-Agent AI for Swine Disease Detection"><span class="ltx_text ltx_ref_tag">2(a)</span></a> presents the confidence distribution of each model, highlighting differences in prediction stability. Models with higher peak densities exhibit more concentrated confidence levels, whereas broader distributions indicate greater variability.</p> </div> <div class="ltx_para ltx_noindent" id="S5.SS2.p3"> <p class="ltx_p" id="S5.SS2.p3.1">Among the evaluated models, GPT-4o demonstrates the most balanced confidence distribution, striking an ideal middle ground between predictive consistency and adaptability. It maintains a moderate peak density with a controlled spread, ensuring stable and well-calibrated confidence scores across different predictions. o1-mini follows closely, exhibiting a slightly broader spread, which suggests higher variability but still maintains reasonable reliability.</p> </div> <div class="ltx_para ltx_noindent" id="S5.SS2.p4"> <p class="ltx_p" id="S5.SS2.p4.1">In contrast, Gemini has the broadest confidence distribution despite its high peak density, indicating significant variability in confidence scores. This inconsistency may lead to overconfidence in certain cases, as its predictions fluctuate more widely than those of the other models, making its reliability less predictable.</p> </div> <div class="ltx_para ltx_noindent" id="S5.SS2.p5"> <p class="ltx_p" id="S5.SS2.p5.1">Given these characteristics, GPT-4o emerges as the optimal choice, offering the best balance between predictive reliability and adaptability. Its ability to maintain stable confidence levels while minimizing extreme variations makes it well-suited for applications requiring both accuracy and robustness in real-time decision-making.</p> </div> </section> <section class="ltx_subsection" id="S5.SS3"> <h3 class="ltx_title ltx_title_subsection"> <span class="ltx_tag ltx_tag_subsection">5.3 </span>Knowledge Retrieval Results</h3> <div class="ltx_para ltx_noindent" id="S5.SS3.p1"> <p class="ltx_p" id="S5.SS3.p1.1">Table <a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#S5.T4" title="Table 4 ‣ 5.3 Knowledge Retrieval Results ‣ 5 Results ‣ When Pigs Get Sick: Multi-Agent AI for Swine Disease Detection"><span class="ltx_text ltx_ref_tag">4</span></a> presents a side-by-side evaluation of our approach against the baseline across multiple datasets and evaluation metrics. The results demonstrate that our method outperforms the baseline in all measured dimensions. The most notable improvement is in Coherence, which increases from 3.92 to 4.61, followed by Relevance, which rises from 3.19 to 4.15. Accuracy also shows substantial gains, particularly in the Disease domain, where it jumps from 1.89 to 4.14 in the validation set. Similarly, the overall Final Score improves significantly, increasing from <span class="ltx_text ltx_font_bold" id="S5.SS3.p1.1.1">3.31</span> to <span class="ltx_text ltx_font_bold" id="S5.SS3.p1.1.2">4.17</span>, highlighting the robustness and consistency of our enhanced method.</p> </div> <figure class="ltx_table" id="S5.T4"> <div class="ltx_inline-block ltx_align_center ltx_transformed_outer" id="S5.T4.6" style="width:433.6pt;height:97.1pt;vertical-align:-0.8pt;"><span class="ltx_transformed_inner" style="transform:translate(-66.8pt,14.8pt) scale(0.764573607185287,0.764573607185287) ;"> <table class="ltx_tabular ltx_guessed_headers ltx_align_middle" id="S5.T4.6.6"> <tbody class="ltx_tbody"> <tr class="ltx_tr" id="S5.T4.6.6.7.1"> <th class="ltx_td ltx_th ltx_th_row ltx_border_l ltx_border_r ltx_border_t" id="S5.T4.6.6.7.1.1" rowspan="2" style="padding-top:2.5pt;padding-bottom:2.5pt;"></th> <th class="ltx_td ltx_align_left ltx_th ltx_th_row ltx_border_r ltx_border_t" id="S5.T4.6.6.7.1.2" rowspan="2" style="padding-top:2.5pt;padding-bottom:2.5pt;"><span class="ltx_text ltx_font_bold" id="S5.T4.6.6.7.1.2.1">Dataset</span></th> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" colspan="2" id="S5.T4.6.6.7.1.3" style="padding-top:2.5pt;padding-bottom:2.5pt;"><span class="ltx_text ltx_font_bold" id="S5.T4.6.6.7.1.3.1">Accuracy</span></td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" colspan="2" id="S5.T4.6.6.7.1.4" style="padding-top:2.5pt;padding-bottom:2.5pt;"><span class="ltx_text ltx_font_bold" id="S5.T4.6.6.7.1.4.1">Relevance</span></td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" colspan="2" id="S5.T4.6.6.7.1.5" style="padding-top:2.5pt;padding-bottom:2.5pt;"><span class="ltx_text ltx_font_bold" id="S5.T4.6.6.7.1.5.1">Correctness</span></td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" colspan="2" id="S5.T4.6.6.7.1.6" style="padding-top:2.5pt;padding-bottom:2.5pt;"><span class="ltx_text ltx_font_bold" id="S5.T4.6.6.7.1.6.1">Coherence</span></td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" colspan="2" id="S5.T4.6.6.7.1.7" style="padding-top:2.5pt;padding-bottom:2.5pt;"><span class="ltx_text ltx_font_bold" id="S5.T4.6.6.7.1.7.1">Expansiveness</span></td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" colspan="2" id="S5.T4.6.6.7.1.8" style="padding-top:2.5pt;padding-bottom:2.5pt;"><span class="ltx_text ltx_font_bold" id="S5.T4.6.6.7.1.8.1">Final Score</span></td> </tr> <tr class="ltx_tr" id="S5.T4.6.6.8.2"> <td class="ltx_td ltx_align_center ltx_border_t" id="S5.T4.6.6.8.2.1" style="padding-top:2.5pt;padding-bottom:2.5pt;">Baseline</td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T4.6.6.8.2.2" style="padding-top:2.5pt;padding-bottom:2.5pt;">Ours</td> <td class="ltx_td ltx_align_center ltx_border_t" id="S5.T4.6.6.8.2.3" style="padding-top:2.5pt;padding-bottom:2.5pt;">Baseline</td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T4.6.6.8.2.4" style="padding-top:2.5pt;padding-bottom:2.5pt;">Ours</td> <td class="ltx_td ltx_align_center ltx_border_t" id="S5.T4.6.6.8.2.5" style="padding-top:2.5pt;padding-bottom:2.5pt;">Baseline</td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T4.6.6.8.2.6" style="padding-top:2.5pt;padding-bottom:2.5pt;">Ours</td> <td class="ltx_td ltx_align_center ltx_border_t" id="S5.T4.6.6.8.2.7" style="padding-top:2.5pt;padding-bottom:2.5pt;">Baseline</td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T4.6.6.8.2.8" style="padding-top:2.5pt;padding-bottom:2.5pt;">Ours</td> <td class="ltx_td ltx_align_center ltx_border_t" id="S5.T4.6.6.8.2.9" style="padding-top:2.5pt;padding-bottom:2.5pt;">Baseline</td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T4.6.6.8.2.10" style="padding-top:2.5pt;padding-bottom:2.5pt;">Ours</td> <td class="ltx_td ltx_align_center ltx_border_t" id="S5.T4.6.6.8.2.11" style="padding-top:2.5pt;padding-bottom:2.5pt;">Baseline</td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T4.6.6.8.2.12" style="padding-top:2.5pt;padding-bottom:2.5pt;">Ours</td> </tr> <tr class="ltx_tr" id="S5.T4.6.6.9.3"> <th class="ltx_td ltx_align_left ltx_th ltx_th_row ltx_border_l ltx_border_r ltx_border_t" id="S5.T4.6.6.9.3.1" rowspan="2" style="padding-top:2.5pt;padding-bottom:2.5pt;"><span class="ltx_text ltx_font_bold" id="S5.T4.6.6.9.3.1.1">Val</span></th> <th class="ltx_td ltx_align_left ltx_th ltx_th_row ltx_border_r ltx_border_t" id="S5.T4.6.6.9.3.2" style="padding-top:2.5pt;padding-bottom:2.5pt;">Vaccine</th> <td class="ltx_td ltx_align_center ltx_border_t" id="S5.T4.6.6.9.3.3" style="padding-top:2.5pt;padding-bottom:2.5pt;">4.12</td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T4.6.6.9.3.4" style="padding-top:2.5pt;padding-bottom:2.5pt;">4.46</td> <td class="ltx_td ltx_align_center ltx_border_t" id="S5.T4.6.6.9.3.5" style="padding-top:2.5pt;padding-bottom:2.5pt;">4.22</td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T4.6.6.9.3.6" style="padding-top:2.5pt;padding-bottom:2.5pt;">4.49</td> <td class="ltx_td ltx_align_center ltx_border_t" id="S5.T4.6.6.9.3.7" style="padding-top:2.5pt;padding-bottom:2.5pt;">4.15</td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T4.6.6.9.3.8" style="padding-top:2.5pt;padding-bottom:2.5pt;">4.47</td> <td class="ltx_td ltx_align_center ltx_border_t" id="S5.T4.6.6.9.3.9" style="padding-top:2.5pt;padding-bottom:2.5pt;">4.51</td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T4.6.6.9.3.10" style="padding-top:2.5pt;padding-bottom:2.5pt;">4.74</td> <td class="ltx_td ltx_align_center ltx_border_t" id="S5.T4.6.6.9.3.11" style="padding-top:2.5pt;padding-bottom:2.5pt;">4.03</td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T4.6.6.9.3.12" style="padding-top:2.5pt;padding-bottom:2.5pt;">4.17</td> <td class="ltx_td ltx_align_center ltx_border_t" id="S5.T4.6.6.9.3.13" style="padding-top:2.5pt;padding-bottom:2.5pt;">4.20</td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T4.6.6.9.3.14" style="padding-top:2.5pt;padding-bottom:2.5pt;">4.46</td> </tr> <tr class="ltx_tr" id="S5.T4.6.6.10.4"> <th class="ltx_td ltx_align_left ltx_th ltx_th_row ltx_border_r" id="S5.T4.6.6.10.4.1" style="padding-top:2.5pt;padding-bottom:2.5pt;">Disease</th> <td class="ltx_td ltx_align_center" id="S5.T4.6.6.10.4.2" style="padding-top:2.5pt;padding-bottom:2.5pt;">1.89</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T4.6.6.10.4.3" style="padding-top:2.5pt;padding-bottom:2.5pt;">4.14</td> <td class="ltx_td ltx_align_center" id="S5.T4.6.6.10.4.4" style="padding-top:2.5pt;padding-bottom:2.5pt;">1.57</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T4.6.6.10.4.5" style="padding-top:2.5pt;padding-bottom:2.5pt;">4.21</td> <td class="ltx_td ltx_align_center" id="S5.T4.6.6.10.4.6" style="padding-top:2.5pt;padding-bottom:2.5pt;">2.09</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T4.6.6.10.4.7" style="padding-top:2.5pt;padding-bottom:2.5pt;">4.15</td> <td class="ltx_td ltx_align_center" id="S5.T4.6.6.10.4.8" style="padding-top:2.5pt;padding-bottom:2.5pt;">2.98</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T4.6.6.10.4.9" style="padding-top:2.5pt;padding-bottom:2.5pt;">4.69</td> <td class="ltx_td ltx_align_center" id="S5.T4.6.6.10.4.10" style="padding-top:2.5pt;padding-bottom:2.5pt;">1.82</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T4.6.6.10.4.11" style="padding-top:2.5pt;padding-bottom:2.5pt;">3.44</td> <td class="ltx_td ltx_align_center" id="S5.T4.6.6.10.4.12" style="padding-top:2.5pt;padding-bottom:2.5pt;">2.06</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T4.6.6.10.4.13" style="padding-top:2.5pt;padding-bottom:2.5pt;">4.13</td> </tr> <tr class="ltx_tr" id="S5.T4.6.6.11.5"> <th class="ltx_td ltx_align_left ltx_th ltx_th_row ltx_border_l ltx_border_r ltx_border_t" id="S5.T4.6.6.11.5.1" rowspan="2" style="padding-top:2.5pt;padding-bottom:2.5pt;"><span class="ltx_text ltx_font_bold" id="S5.T4.6.6.11.5.1.1">Test</span></th> <th class="ltx_td ltx_align_left ltx_th ltx_th_row ltx_border_r ltx_border_t" id="S5.T4.6.6.11.5.2" style="padding-top:2.5pt;padding-bottom:2.5pt;">Vaccine</th> <td class="ltx_td ltx_align_center ltx_border_t" id="S5.T4.6.6.11.5.3" style="padding-top:2.5pt;padding-bottom:2.5pt;">3.50</td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T4.6.6.11.5.4" style="padding-top:2.5pt;padding-bottom:2.5pt;">4.04</td> <td class="ltx_td ltx_align_center ltx_border_t" id="S5.T4.6.6.11.5.5" style="padding-top:2.5pt;padding-bottom:2.5pt;">3.62</td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T4.6.6.11.5.6" style="padding-top:2.5pt;padding-bottom:2.5pt;">4.04</td> <td class="ltx_td ltx_align_center ltx_border_t" id="S5.T4.6.6.11.5.7" style="padding-top:2.5pt;padding-bottom:2.5pt;">3.47</td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T4.6.6.11.5.8" style="padding-top:2.5pt;padding-bottom:2.5pt;">4.04</td> <td class="ltx_td ltx_align_center ltx_border_t" id="S5.T4.6.6.11.5.9" style="padding-top:2.5pt;padding-bottom:2.5pt;">4.10</td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T4.6.6.11.5.10" style="padding-top:2.5pt;padding-bottom:2.5pt;">4.49</td> <td class="ltx_td ltx_align_center ltx_border_t" id="S5.T4.6.6.11.5.11" style="padding-top:2.5pt;padding-bottom:2.5pt;">3.27</td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T4.6.6.11.5.12" style="padding-top:2.5pt;padding-bottom:2.5pt;">3.80</td> <td class="ltx_td ltx_align_center ltx_border_t" id="S5.T4.6.6.11.5.13" style="padding-top:2.5pt;padding-bottom:2.5pt;">3.57</td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T4.6.6.11.5.14" style="padding-top:2.5pt;padding-bottom:2.5pt;">4.08</td> </tr> <tr class="ltx_tr" id="S5.T4.6.6.12.6"> <th class="ltx_td ltx_align_left ltx_th ltx_th_row ltx_border_r" id="S5.T4.6.6.12.6.1" style="padding-top:2.5pt;padding-bottom:2.5pt;">Disease</th> <td class="ltx_td ltx_align_center" id="S5.T4.6.6.12.6.2" style="padding-top:2.5pt;padding-bottom:2.5pt;">3.32</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T4.6.6.12.6.3" style="padding-top:2.5pt;padding-bottom:2.5pt;">3.84</td> <td class="ltx_td ltx_align_center" id="S5.T4.6.6.12.6.4" style="padding-top:2.5pt;padding-bottom:2.5pt;">3.36</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T4.6.6.12.6.5" style="padding-top:2.5pt;padding-bottom:2.5pt;">3.85</td> <td class="ltx_td ltx_align_center" id="S5.T4.6.6.12.6.6" style="padding-top:2.5pt;padding-bottom:2.5pt;">3.42</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T4.6.6.12.6.7" style="padding-top:2.5pt;padding-bottom:2.5pt;">3.83</td> <td class="ltx_td ltx_align_center" id="S5.T4.6.6.12.6.8" style="padding-top:2.5pt;padding-bottom:2.5pt;">4.07</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T4.6.6.12.6.9" style="padding-top:2.5pt;padding-bottom:2.5pt;">4.52</td> <td class="ltx_td ltx_align_center" id="S5.T4.6.6.12.6.10" style="padding-top:2.5pt;padding-bottom:2.5pt;">2.76</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T4.6.6.12.6.11" style="padding-top:2.5pt;padding-bottom:2.5pt;">3.42</td> <td class="ltx_td ltx_align_center" id="S5.T4.6.6.12.6.12" style="padding-top:2.5pt;padding-bottom:2.5pt;">3.39</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T4.6.6.12.6.13" style="padding-top:2.5pt;padding-bottom:2.5pt;">4.02</td> </tr> <tr class="ltx_tr" id="S5.T4.6.6.6" style="background-color:#E6E6E6;"> <th class="ltx_td ltx_align_center ltx_th ltx_th_row ltx_border_b ltx_border_l ltx_border_r ltx_border_t" colspan="2" id="S5.T4.6.6.6.7" style="padding-top:2.5pt;padding-bottom:2.5pt;"><span class="ltx_text ltx_font_bold" id="S5.T4.6.6.6.7.1" style="background-color:#E6E6E6;">Average</span></th> <td class="ltx_td ltx_align_center ltx_border_b ltx_border_t" id="S5.T4.6.6.6.8" style="padding-top:2.5pt;padding-bottom:2.5pt;"><span class="ltx_text" id="S5.T4.6.6.6.8.1" style="background-color:#E6E6E6;">3.21</span></td> <td class="ltx_td ltx_align_center ltx_border_b ltx_border_r ltx_border_t" id="S5.T4.1.1.1.1" style="padding-top:2.5pt;padding-bottom:2.5pt;"><span class="ltx_text ltx_font_bold" id="S5.T4.1.1.1.1.1" style="background-color:#E6E6E6;">4.12<sup class="ltx_sup" id="S5.T4.1.1.1.1.1.1"><span class="ltx_text ltx_font_medium" id="S5.T4.1.1.1.1.1.1.1">†</span></sup></span></td> <td class="ltx_td ltx_align_center ltx_border_b ltx_border_t" id="S5.T4.6.6.6.9" style="padding-top:2.5pt;padding-bottom:2.5pt;"><span class="ltx_text" id="S5.T4.6.6.6.9.1" style="background-color:#E6E6E6;">3.19</span></td> <td class="ltx_td ltx_align_center ltx_border_b ltx_border_r ltx_border_t" id="S5.T4.2.2.2.2" style="padding-top:2.5pt;padding-bottom:2.5pt;"><span class="ltx_text ltx_font_bold" id="S5.T4.2.2.2.2.1" style="background-color:#E6E6E6;">4.15<sup class="ltx_sup" id="S5.T4.2.2.2.2.1.1"><span class="ltx_text ltx_font_medium" id="S5.T4.2.2.2.2.1.1.1">†</span></sup></span></td> <td class="ltx_td ltx_align_center ltx_border_b ltx_border_t" id="S5.T4.6.6.6.10" style="padding-top:2.5pt;padding-bottom:2.5pt;"><span class="ltx_text" id="S5.T4.6.6.6.10.1" style="background-color:#E6E6E6;">3.28</span></td> <td class="ltx_td ltx_align_center ltx_border_b ltx_border_r ltx_border_t" id="S5.T4.3.3.3.3" style="padding-top:2.5pt;padding-bottom:2.5pt;"><span class="ltx_text ltx_font_bold" id="S5.T4.3.3.3.3.1" style="background-color:#E6E6E6;">4.12<sup class="ltx_sup" id="S5.T4.3.3.3.3.1.1"><span class="ltx_text ltx_font_medium" id="S5.T4.3.3.3.3.1.1.1">†</span></sup></span></td> <td class="ltx_td ltx_align_center ltx_border_b ltx_border_t" id="S5.T4.6.6.6.11" style="padding-top:2.5pt;padding-bottom:2.5pt;"><span class="ltx_text" id="S5.T4.6.6.6.11.1" style="background-color:#E6E6E6;">3.92</span></td> <td class="ltx_td ltx_align_center ltx_border_b ltx_border_r ltx_border_t" id="S5.T4.4.4.4.4" style="padding-top:2.5pt;padding-bottom:2.5pt;"><span class="ltx_text ltx_font_bold" id="S5.T4.4.4.4.4.1" style="background-color:#E6E6E6;">4.61<sup class="ltx_sup" id="S5.T4.4.4.4.4.1.1"><span class="ltx_text ltx_font_medium" id="S5.T4.4.4.4.4.1.1.1">†</span></sup></span></td> <td class="ltx_td ltx_align_center ltx_border_b ltx_border_t" id="S5.T4.6.6.6.12" style="padding-top:2.5pt;padding-bottom:2.5pt;"><span class="ltx_text" id="S5.T4.6.6.6.12.1" style="background-color:#E6E6E6;">2.97</span></td> <td class="ltx_td ltx_align_center ltx_border_b ltx_border_r ltx_border_t" id="S5.T4.5.5.5.5" style="padding-top:2.5pt;padding-bottom:2.5pt;"><span class="ltx_text ltx_font_bold" id="S5.T4.5.5.5.5.1" style="background-color:#E6E6E6;">3.71<sup class="ltx_sup" id="S5.T4.5.5.5.5.1.1"><span class="ltx_text ltx_font_medium" id="S5.T4.5.5.5.5.1.1.1">†</span></sup></span></td> <td class="ltx_td ltx_align_center ltx_border_b ltx_border_t" id="S5.T4.6.6.6.13" style="padding-top:2.5pt;padding-bottom:2.5pt;"><span class="ltx_text" id="S5.T4.6.6.6.13.1" style="background-color:#E6E6E6;">3.31</span></td> <td class="ltx_td ltx_align_center ltx_border_b ltx_border_r ltx_border_t" id="S5.T4.6.6.6.6" style="padding-top:2.5pt;padding-bottom:2.5pt;"><span class="ltx_text ltx_font_bold" id="S5.T4.6.6.6.6.1" style="background-color:#E6E6E6;">4.17<sup class="ltx_sup" id="S5.T4.6.6.6.6.1.1"><span class="ltx_text ltx_font_medium" id="S5.T4.6.6.6.6.1.1.1">†</span></sup></span></td> </tr> </tbody> </table> </span></div> <figcaption class="ltx_caption ltx_centering"><span class="ltx_tag ltx_tag_table">Table 4: </span>Comparison of the baseline and our method across different datasets and multiple evaluation metrics.</figcaption> </figure> <div class="ltx_para ltx_noindent" id="S5.SS3.p2"> <p class="ltx_p" id="S5.SS3.p2.1">Table <a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#S5.T5" title="Table 5 ‣ 5.3 Knowledge Retrieval Results ‣ 5 Results ‣ When Pigs Get Sick: Multi-Agent AI for Swine Disease Detection"><span class="ltx_text ltx_ref_tag">5</span></a> presents a statistical comparison of Ours vs. Baseline across five evaluation metrics for both Validation and Test datasets, using paired t-tests. The results indicate statistically significant differences across all metrics, with Accuracy consistently showing the strongest improvements across both datasets. Correctness and Relevance also demonstrate notable improvements. However, Coherence and Expansiveness show contrasting trends: while both exhibit strong improvements in Validation, their statistical significance is weaker in Test, suggesting that the observed differences may be more sensitive to dataset variability. The consistently high t-statistics confirm a meaningful performance difference between our approach and the baseline, though some metrics may generalize better than others.</p> </div> <figure class="ltx_table" id="S5.T5"> <table class="ltx_tabular ltx_centering ltx_guessed_headers ltx_align_middle" id="S5.T5.1"> <thead class="ltx_thead"> <tr class="ltx_tr" id="S5.T5.1.1.1"> <th class="ltx_td ltx_align_left ltx_th ltx_th_column ltx_th_row ltx_border_l ltx_border_r ltx_border_t" id="S5.T5.1.1.1.1" rowspan="2" style="padding:1pt 10.0pt;"><span class="ltx_text ltx_font_bold" id="S5.T5.1.1.1.1.1">Metric</span></th> <th class="ltx_td ltx_align_center ltx_th ltx_th_column ltx_border_r ltx_border_t" colspan="2" id="S5.T5.1.1.1.2" style="padding:1pt 10.0pt;"><span class="ltx_text ltx_font_bold" id="S5.T5.1.1.1.2.1">t-statistic</span></th> <th class="ltx_td ltx_align_center ltx_th ltx_th_column ltx_border_r ltx_border_t" colspan="2" id="S5.T5.1.1.1.3" style="padding:1pt 10.0pt;"><span class="ltx_text ltx_font_bold" id="S5.T5.1.1.1.3.1">p-value</span></th> </tr> <tr class="ltx_tr" id="S5.T5.1.2.2"> <th class="ltx_td ltx_align_center ltx_th ltx_th_column ltx_border_t" id="S5.T5.1.2.2.1" style="padding:1pt 10.0pt;">Validation</th> <th class="ltx_td ltx_align_center ltx_th ltx_th_column ltx_border_r ltx_border_t" id="S5.T5.1.2.2.2" style="padding:1pt 10.0pt;">Test</th> <th class="ltx_td ltx_align_center ltx_th ltx_th_column ltx_border_t" id="S5.T5.1.2.2.3" style="padding:1pt 10.0pt;">Validation</th> <th class="ltx_td ltx_align_center ltx_th ltx_th_column ltx_border_r ltx_border_t" id="S5.T5.1.2.2.4" style="padding:1pt 10.0pt;">Test</th> </tr> </thead> <tbody class="ltx_tbody"> <tr class="ltx_tr" id="S5.T5.1.3.1"> <th class="ltx_td ltx_align_left ltx_th ltx_th_row ltx_border_l ltx_border_r ltx_border_t" id="S5.T5.1.3.1.1" style="padding:1pt 10.0pt;">Accuracy</th> <td class="ltx_td ltx_align_center ltx_border_t" id="S5.T5.1.3.1.2" style="padding:1pt 10.0pt;">4.6787</td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T5.1.3.1.3" style="padding:1pt 10.0pt;">8.3334</td> <td class="ltx_td ltx_align_center ltx_border_t" id="S5.T5.1.3.1.4" style="padding:1pt 10.0pt;">0.0009</td> <td class="ltx_td ltx_align_center ltx_border_r ltx_border_t" id="S5.T5.1.3.1.5" style="padding:1pt 10.0pt;">0.0010</td> </tr> <tr class="ltx_tr" id="S5.T5.1.4.2"> <th class="ltx_td ltx_align_left ltx_th ltx_th_row ltx_border_l ltx_border_r" id="S5.T5.1.4.2.1" style="padding:1pt 10.0pt;">Coherence</th> <td class="ltx_td ltx_align_center" id="S5.T5.1.4.2.2" style="padding:1pt 10.0pt;">4.5289</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T5.1.4.2.3" style="padding:1pt 10.0pt;">6.9511</td> <td class="ltx_td ltx_align_center" id="S5.T5.1.4.2.4" style="padding:1pt 10.0pt;">0.0003</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T5.1.4.2.5" style="padding:1pt 10.0pt;">0.0100</td> </tr> <tr class="ltx_tr" id="S5.T5.1.5.3"> <th class="ltx_td ltx_align_left ltx_th ltx_th_row ltx_border_l ltx_border_r" id="S5.T5.1.5.3.1" style="padding:1pt 10.0pt;">Correctness</th> <td class="ltx_td ltx_align_center" id="S5.T5.1.5.3.2" style="padding:1pt 10.0pt;">3.5681</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T5.1.5.3.3" style="padding:1pt 10.0pt;">7.5433</td> <td class="ltx_td ltx_align_center" id="S5.T5.1.5.3.4" style="padding:1pt 10.0pt;">0.0016</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T5.1.5.3.5" style="padding:1pt 10.0pt;">0.0044</td> </tr> <tr class="ltx_tr" id="S5.T5.1.6.4"> <th class="ltx_td ltx_align_left ltx_th ltx_th_row ltx_border_l ltx_border_r" id="S5.T5.1.6.4.1" style="padding:1pt 10.0pt;">Expansiveness</th> <td class="ltx_td ltx_align_center" id="S5.T5.1.6.4.2" style="padding:1pt 10.0pt;">4.7365</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T5.1.6.4.3" style="padding:1pt 10.0pt;">6.9230</td> <td class="ltx_td ltx_align_center" id="S5.T5.1.6.4.4" style="padding:1pt 10.0pt;">0.0004</td> <td class="ltx_td ltx_align_center ltx_border_r" id="S5.T5.1.6.4.5" style="padding:1pt 10.0pt;">0.0103</td> </tr> <tr class="ltx_tr" id="S5.T5.1.7.5"> <th class="ltx_td ltx_align_left ltx_th ltx_th_row ltx_border_b ltx_border_l ltx_border_r" id="S5.T5.1.7.5.1" style="padding:1pt 10.0pt;">Relevance</th> <td class="ltx_td ltx_align_center ltx_border_b" id="S5.T5.1.7.5.2" style="padding:1pt 10.0pt;">3.5721</td> <td class="ltx_td ltx_align_center ltx_border_b ltx_border_r" id="S5.T5.1.7.5.3" style="padding:1pt 10.0pt;">9.2212</td> <td class="ltx_td ltx_align_center ltx_border_b" id="S5.T5.1.7.5.4" style="padding:1pt 10.0pt;">0.0014</td> <td class="ltx_td ltx_align_center ltx_border_b ltx_border_r" id="S5.T5.1.7.5.5" style="padding:1pt 10.0pt;">0.0063</td> </tr> </tbody> </table> <figcaption class="ltx_caption ltx_centering"><span class="ltx_tag ltx_tag_table">Table 5: </span>Statistical comparison of Ours vs. Baseline on validation and test sets.</figcaption> </figure> <div class="ltx_para ltx_noindent" id="S5.SS3.p3"> <p class="ltx_p" id="S5.SS3.p3.1">To further illustrate how these improvements manifest across different rating scales, Figure <a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#S5.F2.sf2" title="In Figure 2 ‣ 5.2 Symptom-Based Diagnostic Results ‣ 5 Results ‣ When Pigs Get Sick: Multi-Agent AI for Swine Disease Detection"><span class="ltx_text ltx_ref_tag">2(b)</span></a> presents rating distributions. The Coherence category shows a particularly high concentration of top-level ratings, with <span class="ltx_text ltx_font_bold" id="S5.SS3.p3.1.1">77.6%</span> of responses rated 5. In contrast, Expansiveness exhibits a broader distribution, indicating greater variability in response depth. While Correctness, Relevance, and Accuracy also demonstrate strong top-tier ratings, a noticeable proportion of responses still fall within the 3 and 4 range.</p> </div> <div class="ltx_para ltx_noindent" id="S5.SS3.p4"> <p class="ltx_p" id="S5.SS3.p4.1">Overall, these findings confirm that our refinements have significantly enhanced response accuracy, contextual relevance, and the model’s adaptability to real-world, domain-specific queries.</p> </div> </section> </section> <section class="ltx_section" id="S6"> <h2 class="ltx_title ltx_title_section"> <span class="ltx_tag ltx_tag_section">6 </span>Discussion</h2> <div class="ltx_para ltx_noindent" id="S6.p1"> <p class="ltx_p" id="S6.p1.1">The analysis identifies several critical areas contributing to errors in our AI diagnostic system, providing insights into opportunities for further improvement.</p> </div> <section class="ltx_subsection" id="S6.SS1"> <h3 class="ltx_title ltx_title_subsection"> <span class="ltx_tag ltx_tag_subsection">6.1 </span>Ambiguity in Query Classification</h3> <div class="ltx_para ltx_noindent" id="S6.SS1.p1"> <p class="ltx_p" id="S6.SS1.p1.1">Table <a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#S5.T2" title="Table 2 ‣ 5.1 Query Classification Results ‣ 5 Results ‣ When Pigs Get Sick: Multi-Agent AI for Swine Disease Detection"><span class="ltx_text ltx_ref_tag">2</span></a> highlights instances where queries are misclassified, often being incorrectly assigned to the general category instead of their intended classification. These errors often result from ambiguous abbreviations or vague phrasing. For example, queries such as “Please provide usage instructions” and “Ratio” lack sufficient context for precise categorization. Similarly, the query “DLD for Agita” contains the abbreviation DLD, which is intended to refer to the Department of Livestock Development but can be misinterpreted as Digital Learning Development. As a result, the system issues a default response: “Sorry, I cannot provide information on this topic. However, I can assist with swine diseases, vaccine diagnostics, or treatments”. Enhancing context-aware query processing by integrating advanced Natural Language Understanding (NLU) techniques and a targeted Named Entity Recognition (NER) model capable of disambiguating domain-specific abbreviations, such as "DLD," as outlined in Section <a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#S3.SS3" title="3.3 Recommendations ‣ 3 Methodology ‣ When Pigs Get Sick: Multi-Agent AI for Swine Disease Detection"><span class="ltx_text ltx_ref_tag">3.3</span></a>, could significantly improve query classification accuracy.</p> </div> <div class="ltx_para ltx_noindent" id="S6.SS1.p2"> <p class="ltx_p" id="S6.SS1.p2.1">Additionally, five queries have been incorrectly classified under Diagnostic. For instance, the query “Are the breeding pigs infected with African Swine Fever?” includes disease-specific terminology, inadvertently triggering a diagnostic classification, even when the user’s intent is to retrieve general information. Similarly, six retrieval queries have been misclassified under TBC. One example is the query “I encountered dark stool and stool with blood. What injection should I use?”, which prompts the system to seek clarification: “Do you need information on diagnosing or treating dark and bloody stool in pigs? Please confirm”. While these misclassifications are not always critical, they introduce unnecessary friction in the user interaction.</p> </div> </section> <section class="ltx_subsection" id="S6.SS2"> <h3 class="ltx_title ltx_title_subsection"> <span class="ltx_tag ltx_tag_subsection">6.2 </span>Misclassification of Diseases</h3> <div class="ltx_para ltx_noindent" id="S6.SS2.p1"> <p class="ltx_p" id="S6.SS2.p1.1">Table <a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#S5.T3" title="Table 3 ‣ 5.2 Symptom-Based Diagnostic Results ‣ 5 Results ‣ When Pigs Get Sick: Multi-Agent AI for Swine Disease Detection"><span class="ltx_text ltx_ref_tag">3</span></a> reveals strong diagnostic accuracy for some diseases, such as ASF and PED, but greater variability for others, like FMD and PRRS. These inconsistencies may stem from uneven representation in the training data or the inherently subtle symptom profiles of certain diseases.</p> </div> <div class="ltx_para ltx_noindent" id="S6.SS2.p2"> <p class="ltx_p" id="S6.SS2.p2.1">In this case, the actual diagnosis was FMD, but the model sometimes predicted OOD. This could occur because the symptoms were too general and could be linked to other conditions, such as bacterial infections, nutritional deficiencies, or mechanical injuries, making it challenging for the model to precisely identify diseases. Additionally, low model confidence due to missing key diagnostic features may have led it to classify the case as OOD rather than making an uncertain disease assignment. In other instances, the model misclassified FMD as ASF, likely due to overlapping clinical signs such as lethargy and sudden death.</p> </div> <div class="ltx_para ltx_noindent" id="S6.SS2.p3"> <p class="ltx_p" id="S6.SS2.p3.1">Another disease with a low ranking is PRRS. For example, in the case of: “Abortions in sows, an increase in stillborn or mummified fetuses, respiratory symptoms, high fever, red skin, anorexia, and diarrhea”, the symptoms strongly align with PRRS; however, the system misclassified the case as OOD or ASF. The model likely considered the symptoms too broad or overlapping with multiple diseases, leading it to prioritize OOD due to uncertainty. Since ASF, PRRS, and FMD all cause fever, the system failed to recognize the reproductive and respiratory indicators as PRRS-specific. By over-prioritizing fever and redness, the model favored ASF while overlooking the strong reproductive symptoms of PRRS. As a result, PRRS was ranked lower than OOD and ASF, despite its clear symptom alignment.</p> </div> <div class="ltx_para ltx_noindent" id="S6.SS2.p4"> <p class="ltx_p" id="S6.SS2.p4.1">Many swine viral infections share overlapping symptoms, making diagnostic uncertainty a significant challenge. Symptoms such as fever, anorexia, diarrhea, and respiratory distress are common across multiple diseases, often resulting in misclassification. Additionally, because some diseases in the test set had very few cases (e.g., 1/1 or 5/5), the model’s performance for these diseases may not fully reflect its real-world generalization ability. This highlights the need for a more balanced dataset, particularly for underrepresented diseases, to provide a fairer and more robust evaluation of the model’s diagnostic capabilities.</p> </div> </section> <section class="ltx_subsection" id="S6.SS3"> <h3 class="ltx_title ltx_title_subsection"> <span class="ltx_tag ltx_tag_subsection">6.3 </span>Low Confidence Predictions</h3> <div class="ltx_para ltx_noindent" id="S6.SS3.p1"> <p class="ltx_p" id="S6.SS3.p1.1">Despite overall improvements, some predictions received low scores (see Table <a class="ltx_ref" href="https://arxiv.org/html/2503.15204v1#S5.T4" title="Table 4 ‣ 5.3 Knowledge Retrieval Results ‣ 5 Results ‣ When Pigs Get Sick: Multi-Agent AI for Swine Disease Detection"><span class="ltx_text ltx_ref_tag">4</span></a>), indicating areas where the system struggled with alignment, relevance, or completeness. Although fairness in evaluation is essential, a broader concern is the need for comprehensive responses fully capturing clinically relevant details.</p> </div> <div class="ltx_para ltx_noindent" id="S6.SS3.p2"> <p class="ltx_p" id="S6.SS3.p2.1"><span class="ltx_text ltx_font_bold" id="S6.SS3.p2.1.1">Example 1:</span> ASF Testing (Rating: 2)</p> <ul class="ltx_itemize" id="S6.I1"> <li class="ltx_item" id="S6.I1.i1" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item">•</span> <div class="ltx_para" id="S6.I1.i1.p1"> <p class="ltx_p" id="S6.I1.i1.p1.1"><span class="ltx_text ltx_font_bold" id="S6.I1.i1.p1.1.1">Query:</span> “How does a lab test for ASF (African Swine Fever)?”</p> </div> </li> <li class="ltx_item" id="S6.I1.i2" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item">•</span> <div class="ltx_para" id="S6.I1.i2.p1"> <p class="ltx_p" id="S6.I1.i2.p1.1"><span class="ltx_text ltx_font_bold" id="S6.I1.i2.p1.1.1">Prediction:</span> A general overview of ASF diagnostic methods, mentioning PCR and ELISA.</p> </div> </li> <li class="ltx_item" id="S6.I1.i3" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item">•</span> <div class="ltx_para" id="S6.I1.i3.p1"> <p class="ltx_p" id="S6.I1.i3.p1.1"><span class="ltx_text ltx_font_bold" id="S6.I1.i3.p1.1.1">Ground Truth:</span> Detailed laboratory setup, including equipment and procedural steps for ASF testing.</p> </div> </li> <li class="ltx_item" id="S6.I1.i4" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item">•</span> <div class="ltx_para ltx_noindent" id="S6.I1.i4.p1"> <p class="ltx_p" id="S6.I1.i4.p1.1"><span class="ltx_text ltx_font_bold" id="S6.I1.i4.p1.1.1">Issue:</span> While the response correctly identifies PCR and ELISA as valid diagnostic methods, it lacks the detailed procedural and equipment-related information found in the reference. The lower score reflects incompleteness rather than inaccuracy, as the response omits but does not contradict essential details.</p> </div> </li> </ul> </div> <div class="ltx_para ltx_noindent" id="S6.SS3.p3"> <p class="ltx_p" id="S6.SS3.p3.1"><span class="ltx_text ltx_font_bold" id="S6.SS3.p3.1.1">Example 2:</span> Water Disinfection (Rating: 1)</p> <ul class="ltx_itemize" id="S6.I2"> <li class="ltx_item" id="S6.I2.i1" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item">•</span> <div class="ltx_para" id="S6.I2.i1.p1"> <p class="ltx_p" id="S6.I2.i1.p1.1"><span class="ltx_text ltx_font_bold" id="S6.I2.i1.p1.1.1">Query:</span> “How to disinfect water before use?”</p> </div> </li> <li class="ltx_item" id="S6.I2.i2" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item">•</span> <div class="ltx_para" id="S6.I2.i2.p1"> <p class="ltx_p" id="S6.I2.i2.p1.1"><span class="ltx_text ltx_font_bold" id="S6.I2.i2.p1.1.1">Prediction:</span> Describes the use of hydrogen peroxide for water disinfection.</p> </div> </li> <li class="ltx_item" id="S6.I2.i3" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item">•</span> <div class="ltx_para" id="S6.I2.i3.p1"> <p class="ltx_p" id="S6.I2.i3.p1.1"><span class="ltx_text ltx_font_bold" id="S6.I2.i3.p1.1.1">Ground Truth:</span> Specifies chlorine-based disinfectants and correct dosage instructions.</p> </div> </li> <li class="ltx_item" id="S6.I2.i4" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item">•</span> <div class="ltx_para ltx_noindent" id="S6.I2.i4.p1"> <p class="ltx_p" id="S6.I2.i4.p1.1"><span class="ltx_text ltx_font_bold" id="S6.I2.i4.p1.1.1">Issue:</span> The Prediction presents a scientifically valid but alternative method not covered in the reference. The misalignment stems from the Ground Truth’s exclusive focus on chlorine-based disinfection, despite hydrogen peroxide being a legitimate option. The lower score reflects a lack of alignment rather than factual inaccuracy.</p> </div> </li> </ul> </div> <div class="ltx_para ltx_noindent" id="S6.SS3.p4"> <p class="ltx_p" id="S6.SS3.p4.1"><span class="ltx_text ltx_font_bold" id="S6.SS3.p4.1.1">Example 3:</span> Pig Diarrhea Treatment (Rating: 0.5)</p> <ul class="ltx_itemize" id="S6.I3"> <li class="ltx_item" id="S6.I3.i1" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item">•</span> <div class="ltx_para" id="S6.I3.i1.p1"> <p class="ltx_p" id="S6.I3.i1.p1.1"><span class="ltx_text ltx_font_bold" id="S6.I3.i1.p1.1.1">Query:</span> “Which drug group can be used to treat black diarrhea in pigs?”</p> </div> </li> <li class="ltx_item" id="S6.I3.i2" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item">•</span> <div class="ltx_para" id="S6.I3.i2.p1"> <p class="ltx_p" id="S6.I3.i2.p1.1"><span class="ltx_text ltx_font_bold" id="S6.I3.i2.p1.1.1">Prediction:</span> The system recommended Tiamulin as the treatment.</p> </div> </li> <li class="ltx_item" id="S6.I3.i3" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item">•</span> <div class="ltx_para" id="S6.I3.i3.p1"> <p class="ltx_p" id="S6.I3.i3.p1.1"><span class="ltx_text ltx_font_bold" id="S6.I3.i3.p1.1.1">Ground Truth:</span> The reference document specifies Lincomycin and Spectinomycin as the correct treatment, along with dosage instructions.</p> </div> </li> <li class="ltx_item" id="S6.I3.i4" style="list-style-type:none;"> <span class="ltx_tag ltx_tag_item">•</span> <div class="ltx_para ltx_noindent" id="S6.I3.i4.p1"> <p class="ltx_p" id="S6.I3.i4.p1.1"><span class="ltx_text ltx_font_bold" id="S6.I3.i4.p1.1.1">Issue:</span> The Prediction suggests Tiamulin, a valid treatment not mentioned in the Ground Truth. While the reference lists Lincomycin and Spectinomycin, it does not explicitly exclude other treatments. The lower score reflects a mismatch with the reference rather than an actual error.</p> </div> </li> </ul> </div> <div class="ltx_para ltx_noindent" id="S6.SS3.p5"> <p class="ltx_p" id="S6.SS3.p5.1">Analysis of the evaluation results indicates that certain factually correct yet alternative responses received lower scores primarily due to differences from the reference rather than genuine inaccuracies. This suggests that strict adherence to reference documents sometimes outweighs the recognition of clinically valid variations. While maintaining evaluation consistency remains important, responses should not only reflect alignment with reference materials but also emphasize completeness and clinical applicability. To address this, the knowledge retrieval component should incorporate a broader range of validated sources and expert-vetted alternatives, provide detailed procedural explanations, and recognize multiple clinically valid approaches. This strategy would enhance the depth and practical relevance of responses, ensuring scientific accuracy without overly restricting them to a single reference source.</p> </div> </section> </section> <section class="ltx_section" id="S7"> <h2 class="ltx_title ltx_title_section"> <span class="ltx_tag ltx_tag_section">7 </span>Conclusion and Future works</h2> <div class="ltx_para ltx_noindent" id="S7.p1"> <p class="ltx_p" id="S7.p1.1">This study demonstrates the robust capabilities of an AI-powered multi-agent diagnostic system designed to improve swine disease surveillance and veterinary decision-making. The system effectively distinguishes between different types of user queries—general, retrieval, diagnostic, and ambiguous—and shows notable strengths in diagnostic accuracy and response speed, particularly using the GPT-4o model. The knowledge retrieval module has also significantly advanced in accuracy, coherence, and practical applicability, surpassing previous methods.</p> </div> <div class="ltx_para ltx_noindent" id="S7.p2"> <p class="ltx_p" id="S7.p2.1">However, certain areas remain open for further refinement. The overlap of clinical symptoms among swine diseases occasionally leads to diagnostic misclassification, suggesting that differential diagnostic capabilities can be further enhanced by improving symptom extraction precision and employing multimodal symptom analysis. Additionally, the system currently faces challenges when processing ambiguous or incomplete user queries. This underscores the need for more robust context-aware query interpretation mechanisms, better interactive clarification processes, and more sophisticated entity recognition models.</p> </div> <div class="ltx_para ltx_noindent" id="S7.p3"> <p class="ltx_p" id="S7.p3.1">Improving response expansiveness and relevance is another critical direction for future research. Incorporating comprehensive explanations that include detailed procedural steps, multiple valid treatment options, and extensive clinical contexts will enhance the practical utility of generated responses. Establishing iterative expert-driven feedback loops will further ensure continuous improvement in accuracy and practical applicability.</p> </div> <div class="ltx_para ltx_noindent" id="S7.p4"> <p class="ltx_p" id="S7.p4.1">To enhance real-time practical usability, optimizing the model inference efficiency through techniques such as compression, quantization, or streamlined architectures is also essential. Future work will additionally prioritize expanding the scope of disease coverage, particularly to address emerging and region-specific swine diseases. Integrating comprehensive veterinary procedural databases will further enhance the system’s practical utility by providing detailed, step-by-step guidance for veterinary procedures.</p> </div> <div class="ltx_para ltx_noindent" id="S7.p5"> <p class="ltx_p" id="S7.p5.1">These refinements will further strengthen the system’s effectiveness, generalization capability, and practical usability, thereby significantly advancing veterinary medicine, animal health outcomes, and global agricultural sustainability.</p> </div> </section> <section class="ltx_section" id="S8"> <h2 class="ltx_title ltx_title_section"> <span class="ltx_tag ltx_tag_section">8 </span>Acknowledgments</h2> <div class="ltx_para ltx_noindent" id="S8.p1"> <p class="ltx_p" id="S8.p1.1">We sincerely thank the Swine Veterinary Service (SVS) team for identifying key pain points and providing valuable consultation, which helped shape the direction of this research. We are also grateful to SVS for curating the dataset and ensuring high-quality data selection and expert annotation by veterinary professionals.</p> </div> <div class="ltx_para ltx_noindent" id="S8.p2"> <p class="ltx_p" id="S8.p2.1">Our gratitude extends to our NLP team, Aunchana Pimpisal and Punnawit Foithong, for their insightful discussions and feedback, as well as to the Axons Farm Business team for their efforts in translating this research into real-world applications for future improvements.</p> </div> <div class="ltx_para ltx_noindent" id="S8.p3"> <p class="ltx_p" id="S8.p3.1">Finally, we appreciate Charoen Pokphand Foods (CPF) for initiating CPF Hack the Future, which brought us together—from the initial project pitch to this meaningful collaboration.</p> </div> </section> <section class="ltx_bibliography" id="bib"> <h2 class="ltx_title ltx_title_bibliography">References</h2> <ul class="ltx_biblist"> <li class="ltx_bibitem" id="bib.bib1"> <span class="ltx_tag ltx_tag_bibitem">[1]</span> <span class="ltx_bibblock"> Kimberly VanderWaal and John Deen. </span> <span class="ltx_bibblock">Global trends in infectious diseases of swine. </span> <span class="ltx_bibblock"><span class="ltx_text ltx_font_italic" id="bib.bib1.1.1">Proceedings of the National Academy of Sciences (PNAS)</span>, 115(45):11495–11500, 2018. </span> </li> <li class="ltx_bibitem" id="bib.bib2"> <span class="ltx_tag ltx_tag_bibitem">[2]</span> <span class="ltx_bibblock"> Yubin Kim, Chanwoo Park, Hyewon Jeong, Yik Siu Chan, Xuhai Xu, Daniel McDuff, Hyeonhoon Lee, Marzyeh Ghassemi, Cynthia Breazeal, and Hae Won Park. </span> <span class="ltx_bibblock">Mdagents: An adaptive collaboration of llms for medical decision-making. </span> <span class="ltx_bibblock">In <span class="ltx_text ltx_font_italic" id="bib.bib2.1.1">Proceedings of the 38th Conference on Neural Information Processing Systems (NeurIPS)</span>, 2024. </span> </li> <li class="ltx_bibitem" id="bib.bib3"> <span class="ltx_tag ltx_tag_bibitem">[3]</span> <span class="ltx_bibblock"> Haochun Wang, Sendong Zhao, Zewen Qiang, Nuwa Xi, Bing Qin, and Ting Liu. </span> <span class="ltx_bibblock">Beyond direct diagnosis: Llm-based multi-specialist agent consultation for automatic diagnosis, 2024. </span> </li> <li class="ltx_bibitem" id="bib.bib4"> <span class="ltx_tag ltx_tag_bibitem">[4]</span> <span class="ltx_bibblock"> Thanveer Shaik, Xiaohui Tao, Lin Li, Haoran Xie, Hong-Ning Dai, Feng Zhao, and Jianming Yong. </span> <span class="ltx_bibblock">Adaptive multi-agent deep reinforcement learning for timely healthcare interventions, 2023. </span> </li> <li class="ltx_bibitem" id="bib.bib5"> <span class="ltx_tag ltx_tag_bibitem">[5]</span> <span class="ltx_bibblock"> Sidian Lin, Soroush Saghafian, Jessica M. Lipschitz, and Katherine E. Burdick. </span> <span class="ltx_bibblock">A multi-agent reinforcement learning algorithm for personalized recommendations in bipolar disorder. </span> <span class="ltx_bibblock"><span class="ltx_text ltx_font_italic" id="bib.bib5.1.1">PNAS Nexus</span>, 2024. </span> </li> <li class="ltx_bibitem" id="bib.bib6"> <span class="ltx_tag ltx_tag_bibitem">[6]</span> <span class="ltx_bibblock"> Rupinder Kaur Walia and Harjot Kaur. </span> <span class="ltx_bibblock">Multi-agent-based ensemble learning model with feature selection for enhanced covid-19 detection. </span> <span class="ltx_bibblock"><span class="ltx_text ltx_font_italic" id="bib.bib6.1.1">Journal of Communicable Diseases</span>, 56(1):83–98, 2024. </span> </li> <li class="ltx_bibitem" id="bib.bib7"> <span class="ltx_tag ltx_tag_bibitem">[7]</span> <span class="ltx_bibblock"> Yu He Ke, Rui Yang, Sui An Lie, Taylor Xin Yi Lim, Hairil Rizal Abdullah, Daniel Shu Wei Ting, and Nan Liu. </span> <span class="ltx_bibblock">Enhancing diagnostic accuracy through multi-agent conversations: Using large language models to mitigate cognitive bias, 2024. </span> </li> <li class="ltx_bibitem" id="bib.bib8"> <span class="ltx_tag ltx_tag_bibitem">[8]</span> <span class="ltx_bibblock"> K. E. Sobkowich. </span> <span class="ltx_bibblock">Demystifying artificial intelligence for veterinary professionals: Practical applications and future potential. </span> <span class="ltx_bibblock"><span class="ltx_text ltx_font_italic" id="bib.bib8.1.1">American Journal of Veterinary Research</span>, 56(1):1–10, 2025. </span> </li> <li class="ltx_bibitem" id="bib.bib9"> <span class="ltx_tag ltx_tag_bibitem">[9]</span> <span class="ltx_bibblock"> Olalekan Chris Akinsulie, Ibrahim Idris, Victor Ayodele Aliyu, Sammuel Shahzad, Olamilekan Gabriel Banwo, Seto Charles Ogunleye, Mercy Olorunshola, Deborah O. Okedoyin, Charles Ugwu, Ifeoluwa Peace Oladapo, Joy Olaoluwa Gbadegoye, Qudus Afolabi Akande, Pius Babawale, Sahar Rostami, and Kehinde Olugboyega Soetan. </span> <span class="ltx_bibblock">The potential application of artificial intelligence in veterinary clinical practice and biomedical research. </span> <span class="ltx_bibblock"><span class="ltx_text ltx_font_italic" id="bib.bib9.1.1">Frontiers in Veterinary Science</span>, 11, 2024. </span> </li> <li class="ltx_bibitem" id="bib.bib10"> <span class="ltx_tag ltx_tag_bibitem">[10]</span> <span class="ltx_bibblock"> Steve Joslyn and Kate Alexander. </span> <span class="ltx_bibblock">Evaluating artificial intelligence algorithms for use in veterinary radiology. </span> <span class="ltx_bibblock"><span class="ltx_text ltx_font_italic" id="bib.bib10.1.1">Veterinary Radiology & Ultrasound</span>, 63(Suppl. 1):871–879, 2022. </span> </li> <li class="ltx_bibitem" id="bib.bib11"> <span class="ltx_tag ltx_tag_bibitem">[11]</span> <span class="ltx_bibblock"> Yixing Jiang, Jeremy Irvin, Andrew Y. Ng, and James Zou. </span> <span class="ltx_bibblock">VetLLM: Large language model for predicting diagnosis from veterinary notes. </span> <span class="ltx_bibblock">In <span class="ltx_text ltx_font_italic" id="bib.bib11.1.1">Proceedings of the Pacific Symposium on Biocomputing</span>, volume 29, pages 120–133, 2024. </span> </li> <li class="ltx_bibitem" id="bib.bib12"> <span class="ltx_tag ltx_tag_bibitem">[12]</span> <span class="ltx_bibblock"> Brian Hur, Lucy Lu Wang, Laura Hardefeldt, and Meliha Yetsigen. </span> <span class="ltx_bibblock">Is that the right dose? investigating generative language model performance on veterinary prescription text analysis. </span> <span class="ltx_bibblock">In <span class="ltx_text ltx_font_italic" id="bib.bib12.1.1">Proceedings of the 23rd Workshop on Biomedical Language Processing</span>, pages 390–397, 2024. </span> </li> <li class="ltx_bibitem" id="bib.bib13"> <span class="ltx_tag ltx_tag_bibitem">[13]</span> <span class="ltx_bibblock"> Weihao Yu, Zhengyuan Yang, Linjie Li, Jianfeng Wang, Kevin Lin, Zicheng Liu, Xinchao Wang, and Lijuan Wang. </span> <span class="ltx_bibblock">Mm-vet: Evaluating large multimodal models for integrated capabilities. </span> <span class="ltx_bibblock">In <span class="ltx_text ltx_font_italic" id="bib.bib13.1.1">Proceedings of the 41st International Conference on Machine Learning</span>, volume 235 of <span class="ltx_text ltx_font_italic" id="bib.bib13.2.2">PMLR</span>, 2024. </span> </li> <li class="ltx_bibitem" id="bib.bib14"> <span class="ltx_tag ltx_tag_bibitem">[14]</span> <span class="ltx_bibblock"> Alberto L. Barriuso, Gabriel Villarrubia González, Juan F. De Paz, Álvaro Lozano, and Javier Bajo. </span> <span class="ltx_bibblock">Combination of multi-agent systems and wireless sensor networks for the monitoring of cattle. </span> <span class="ltx_bibblock"><span class="ltx_text ltx_font_italic" id="bib.bib14.1.1">Sensors</span>, 18(1):108, 2018. </span> </li> <li class="ltx_bibitem" id="bib.bib15"> <span class="ltx_tag ltx_tag_bibitem">[15]</span> <span class="ltx_bibblock"> Aliva Bakshi, Jake Stetson, Doina Caragea, Lihua Wang, Jishu Shi, and Laura C. Miller. </span> <span class="ltx_bibblock">Toward a rapid, sensitive, user-friendly, field-deployable artificial intelligence tool for enhancing african swine fever diagnosis and reporting. </span> <span class="ltx_bibblock"><span class="ltx_text ltx_font_italic" id="bib.bib15.1.1">American Journal of Veterinary Research</span>, 2025. </span> </li> <li class="ltx_bibitem" id="bib.bib16"> <span class="ltx_tag ltx_tag_bibitem">[16]</span> <span class="ltx_bibblock"> Eddiemar B. Lagua, Hong-Seok Mun, Keiven Mark B. Ampode, Veasna Chem, Young-Hwa Kim, and Chul-Ju Yang. </span> <span class="ltx_bibblock">Artificial intelligence for automatic monitoring of respiratory health conditions in smart swine farming. </span> <span class="ltx_bibblock"><span class="ltx_text ltx_font_italic" id="bib.bib16.1.1">Animals</span>, 13(1860), 2023. </span> </li> <li class="ltx_bibitem" id="bib.bib17"> <span class="ltx_tag ltx_tag_bibitem">[17]</span> <span class="ltx_bibblock"> Madonna Benjamin and Steven Yik. </span> <span class="ltx_bibblock">Precision livestock farming in swine welfare: A review for swine practitioners. </span> <span class="ltx_bibblock"><span class="ltx_text ltx_font_italic" id="bib.bib17.1.1">Animals</span>, 9(4):133, 2019. </span> </li> <li class="ltx_bibitem" id="bib.bib18"> <span class="ltx_tag ltx_tag_bibitem">[18]</span> <span class="ltx_bibblock"> Matthias Eddicks, Franziska Feicht, Jochen Beckjunker, Marika Genzow, Carmen Alonso, Sven Reese, Mathias Ritzmann, and Julia Stadler. </span> <span class="ltx_bibblock">Monitoring of respiratory disease patterns in a multimicrobially infected pig population using artificial intelligence and aggregate samples. </span> <span class="ltx_bibblock"><span class="ltx_text ltx_font_italic" id="bib.bib18.1.1">Viruses</span>, 16(1575), 2024. </span> </li> <li class="ltx_bibitem" id="bib.bib19"> <span class="ltx_tag ltx_tag_bibitem">[19]</span> <span class="ltx_bibblock"> Harry Wooseuk Ryu and Joo Ho Tai. </span> <span class="ltx_bibblock">Object detection and tracking using a high-performance artificial intelligence-based 3d depth camera: Towards early detection of african swine fever. </span> <span class="ltx_bibblock"><span class="ltx_text ltx_font_italic" id="bib.bib19.1.1">Journal of Veterinary Science</span>, 23(1):e17, 2022. </span> </li> <li class="ltx_bibitem" id="bib.bib20"> <span class="ltx_tag ltx_tag_bibitem">[20]</span> <span class="ltx_bibblock"> Shunli Wang, Honghua Jiang, Yongliang Qiao, Shuzhen Jiang, Huaiqin Lin, and Qian Sun. </span> <span class="ltx_bibblock">The research progress of vision-based artificial intelligence in smart pig farming. </span> <span class="ltx_bibblock"><span class="ltx_text ltx_font_italic" id="bib.bib20.1.1">Sensors</span>, 22(17):6541, 2022. </span> </li> <li class="ltx_bibitem" id="bib.bib21"> <span class="ltx_tag ltx_tag_bibitem">[21]</span> <span class="ltx_bibblock"> Mohamad-Hani Temsah, Amr Jamal, Khalid Alhasan, Abdulkarim A. Temsah, and Khalid H. Malki. </span> <span class="ltx_bibblock">Openai o1-preview vs. chatgpt in healthcare: A new frontier in medical ai reasoning. </span> <span class="ltx_bibblock"><span class="ltx_text ltx_font_italic" id="bib.bib21.1.1">Cureus</span>, 16(10):e70640, 2024. </span> </li> <li class="ltx_bibitem" id="bib.bib22"> <span class="ltx_tag ltx_tag_bibitem">[22]</span> <span class="ltx_bibblock"> Nan Zhang, Zaijie Sun, Yuchen Xie, Haiyang Wu, and Cheng Li. </span> <span class="ltx_bibblock">The latest version chatgpt powered by gpt-4o: what will it bring to the medical field? </span> <span class="ltx_bibblock"><span class="ltx_text ltx_font_italic" id="bib.bib22.1.1">International Journal of Surgery</span>, 110:6018–6019, 2024. </span> </li> <li class="ltx_bibitem" id="bib.bib23"> <span class="ltx_tag ltx_tag_bibitem">[23]</span> <span class="ltx_bibblock"> Khaled Saab, Tao Tu, Wei-Hung Weng, Ryutaro Tanno, David Stutz, and et al. </span> <span class="ltx_bibblock">Capabilities of gemini models in medicine, 2024. </span> </li> <li class="ltx_bibitem" id="bib.bib24"> <span class="ltx_tag ltx_tag_bibitem">[24]</span> <span class="ltx_bibblock"> Google Cloud. </span> <span class="ltx_bibblock">Vertex ai search. </span> <span class="ltx_bibblock"><a class="ltx_ref ltx_url ltx_font_typewriter" href="https://cloud.google.com/vertex-ai" title="">https://cloud.google.com/vertex-ai</a>, 2023. </span> <span class="ltx_bibblock">Generally available in August 2023. Accessed: 2025-02-19. </span> </li> </ul> </section> <div class="ltx_pagination ltx_role_newpage"></div> </article> </div> <footer class="ltx_page_footer"> <div class="ltx_page_logo">Generated on Wed Mar 19 13:42:33 2025 by <a class="ltx_LaTeXML_logo" href="http://dlmf.nist.gov/LaTeXML/"><span style="letter-spacing:-0.2em; margin-right:0.1em;">L<span class="ltx_font_smallcaps" style="position:relative; bottom:2.2pt;">a</span>T<span class="ltx_font_smallcaps" style="font-size:120%;position:relative; bottom:-0.2ex;">e</span></span><span style="font-size:90%; position:relative; bottom:-0.2ex;">XML</span><img alt="Mascot Sammy" src="data:image/png;base64,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"/></a> </div></footer> </div> </body> </html>