[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"detail-sidebar-cat-0-en-105":3,"doc-seo-273366-105":59,"doc-detail-273366-en":131},{"code":4,"msg":5,"data":6},0,"success",[7,13,18,23,28,33,38,43,48,51,55],{"id":8,"doc_module":4,"doc_module_name":9,"category_name":10,"show_sort_weight":11,"slug":12},1,"Document","Story & Novel",90,"story-novel",{"id":14,"doc_module":4,"doc_module_name":9,"category_name":15,"show_sort_weight":16,"slug":17},2,"Literature",80,"literature",{"id":19,"doc_module":4,"doc_module_name":9,"category_name":20,"show_sort_weight":21,"slug":22},4,"Exam",70,"exam",{"id":24,"doc_module":4,"doc_module_name":9,"category_name":25,"show_sort_weight":26,"slug":27},5,"Comic",60,"comic",{"id":29,"doc_module":4,"doc_module_name":9,"category_name":30,"show_sort_weight":31,"slug":32},6,"Technology",50,"technology",{"id":34,"doc_module":4,"doc_module_name":9,"category_name":35,"show_sort_weight":36,"slug":37},7,"Healthcare",40,"healthcare",{"id":39,"doc_module":4,"doc_module_name":9,"category_name":40,"show_sort_weight":41,"slug":42},8,"Research & Report",30,"research-report",{"id":44,"doc_module":4,"doc_module_name":9,"category_name":45,"show_sort_weight":46,"slug":47},9,"Religion & Spirituality",20,"religion-spirituality",{"id":46,"doc_module":4,"doc_module_name":9,"category_name":49,"show_sort_weight":46,"slug":50},"World Cup","world-cup",{"id":52,"doc_module":4,"doc_module_name":9,"category_name":53,"show_sort_weight":52,"slug":54},10,"Lifestyle","lifestyle",{"id":56,"doc_module":4,"doc_module_name":9,"category_name":57,"show_sort_weight":24,"slug":58},19,"General","general",{"code":4,"msg":60,"data":61},"ok",{"site_id":62,"language":63,"slug":64,"title":65,"keywords":66,"description":67,"schema_data":68,"social_meta":124,"head_meta":126,"extra_data":128,"updated_unix":130},105,"en","medragchecker-claim-level-verification-for-biomedical-retrieval-augmented-generation","MedRAGChecker - Claim-Level Verification for Biomedical Retrieval-Augmented Generation","","Biomedical retrieval-augmented generation (RAG) often produces long-form answers that mix supported, under-evidenced, and contradicted statements, making fine-grained verification essential. The work introduces MEDRAGCHECKER, which decomposes answers into atomic claims, verifies each claim via textual NLI plus knowledge-graph (KG) consistency, and fuses signals using selective gating when KG conflict is decisive. Evaluated with distilled GPT-4.1 student checkers across PubMedQA, MedQuAD, LiveQA, and MedRedQA, MEDRAGCHECKER matches RAG faithfulness while adding safety-critical error diagnostics, including decision-flip alignment with human judgment (69.8% vs 63.4% NLI-only).",{"@graph":69,"@context":123},[70,84,106],{"@type":71,"itemListElement":72},"BreadcrumbList",[73,77,79,82],{"item":74,"name":75,"@type":76,"position":8},"https://docshare.wps.com","Home","ListItem",{"item":78,"name":9,"@type":76,"position":14},"https://docshare.wps.com/document/",{"item":80,"name":40,"@type":76,"position":81},"https://docshare.wps.com/document/research-report/",3,{"item":83,"name":65,"@type":76,"position":19},"https://docshare.wps.com/document/medragchecker-claim-level-verification-for-biomedical-retrieval-augmented-generation/273366/",{"url":83,"name":65,"@type":85,"image":86,"author":91,"headline":65,"publisher":94,"fileFormat":97,"inLanguage":63,"description":67,"dateModified":98,"datePublished":99,"encodingFormat":97,"isAccessibleForFree":100,"interactionStatistic":101},"DigitalDocument",{"url":87,"@type":88,"width":89,"height":90},"https://docshare.wps.com/thumbnails/medragchecker-claim-level-verification-for-biomedical-retrieval-augmented-generation/273366.png","ImageObject",300,407,{"name":92,"@type":93},"Logic","Person",{"url":74,"name":95,"@type":96},"DocShare","Organization","application/pdf","2026-11-07","2026-09-15",true,{"@type":102,"interactionType":103,"userInteractionCount":105},"InteractionCounter",{"@type":104},"ViewAction",22,{"@type":107,"mainEntity":108},"FAQPage",[109,115,119],{"name":110,"@type":111,"acceptedAnswer":112},"What problem does MEDRAGCHECKER address in biomedical retrieval-augmented generation?","Question",{"text":113,"@type":114},"Long-form biomedical RAG answers can contain isolated clinically important errors that whole-answer scoring may hide. MEDRAGCHECKER targets this by verifying claims at a finer granularity.","Answer",{"name":116,"@type":111,"acceptedAnswer":117},"How does MEDRAGCHECKER verify biomedical claims?",{"text":118,"@type":114},"It decomposes generated answers into atomic claims, then verifies each claim using textual natural language inference (NLI) combined with a knowledge-graph (KG) consistency signal.",{"name":120,"@type":111,"acceptedAnswer":121},"Why does the method use selective gating instead of always trusting the KG?",{"text":122,"@type":114},"The KG can help or hurt depending on whether it covers the relevant claim entities and relations. Selective gating fuses the KG signal only when it strongly conflicts with a confident NLI entailment.","https://schema.org",{"og:url":83,"og:type":125,"og:title":65,"og:site_name":95,"og:description":67},"article",{"robots":127,"canonical":83},"index,follow",{"doc_id":129,"site_id":62},273366,1789471083,{"code":4,"msg":5,"data":132},{"doc_id":129,"user_id":133,"nickname":92,"user_avatar":134,"doc_module":4,"category_id":39,"category_name":40,"doc_title":65,"doc_description":67,"doc_content":135,"file_id":136,"file_url":137,"file_type":138,"file_size":139,"view_count":105,"is_deleted":4,"is_public":8,"is_downloadable":8,"audit_status":8,"page_count":140,"language":141,"language_code":63,"site_id":62,"html_lang":63,"table_of_contents":142,"faqs":143,"seo_title":144,"seo_description":67,"update_tm":130,"read_time":145,"preview_page_count":4},1099513958762,"https://ap-avatar.wpscdn.com/avatar/1000023916a998db790?x-image-process=image/resize,m_fixed,w_180,h_180&k=1784791008015729253","MedRAGChecker: Claim-Level Verification for Biomedical Retrieval-Augmented Generation  \nYuelyu Ji Min Gu Kwak Hang Zhang Xizhi Wu Chenyu Li Yanshan Wang  \nUniversity of Pittsburgh, Pittsburgh, PA, USA  \n[yueluji@gmail.com](yueluji@gmail.com)  \narXiv : 260 1 .065 19v2 [cs .CL] 2 1 Aug 2026  \nAbstract  \nBiomedical retrieval-augmented generation (RAG) produces long-form answers that mix supported, under-evidenced, and contradicted claims, demanding fine-grained verification. We propose MEDRAGCHECKER, a claimlevel diagnostic framework that (i) decomposes answers into atomic claims,(ii) verifies each claim by combining textual natural language inference (NLI) with a knowledge-graph (KG) consistency signal, and (iii) uses a selective gating rule to fuse the two signals only when the KG signal is decisive, falling back to NLI otherwise. To assess our framework on biomedical RAG, we distill a GPT-4.1 teacher into compact student checkers and evaluate across four biomedical QA benchmarks (PubMedQA, MedQuAD, LiveQA, MedRedQA) with three biomedical generators. MEDRAGCHECKER achieves comparable answer-level faithfulness to general-purpose RAG evaluators while additionally producing safety-critical error diagnostics that align with human judgment on decision-flip claims (69.8% agreement vs.  \n63.4% for NLI-only) . As an additional analysis, we instantiate the KG component with two back-ends of differing scope—DRKG and BioPortal—and find that KG coverage substantially affects fusion quality: an incomplete KG can inflate hallucination via false contradictionson uncovered claims, motivating our selective gating design.  \n1 Introduction  \nLarge language models (LLMs) now pass standardized medical exams, yet their ungrounded generations still produce clinically harmful errors (Yu  \nCode, prompts, and diagnostics available at: [https://](https://)[ ](https://)[github.com/JoyDajunSpaceCraft/MedicalRagChecker](github.com/JoyDajunSpaceCraft/MedicalRagChecker)  \nFigure 1: Selective gating in MEDRAGCHECKER. For the atomic claim “Aspirin is safe for children with viral infections,” the textual NLI checker is over-confident in entailment (pNLI =0 .93) because the retrieved evidence does not mention Reye’s syndrome. The KG verifier returns low support (sKG =0 .05) . The selective gating condition (high pNLI , low sKG ) fires, fusing the two signals into P⋆ =0 .18 and producing the correct NOT ENTAILED verdict that matches the human label.  \net al., 2025 ; Singhal et al., 2025 ; Vladika et al., 2024 ; Pandit et al., 2025) . Retrieval-augmented generation (RAG) partly addresses this by conditioning answers on up-to-date medical literature (Achiam et al., 2023 ; Tu et al., 2024 ; Lewis et al., 2020), but noisy retrieval can still cause the model to anchor on misleading snippets (Ning et al., 2025) . Long-form biomedical answers contain many distinct factual statements, and wholeanswer scoring hides isolated but clinically important mistakes. This motivates claim-level evaluation: decomposing each answer into atomic factual claims (e.g., a single drug–effect or drug– contraindication assertion) and verifying each one independently (Min et al., 2023 ; Ru et al., 2024) .  \nStandard claim-level verification frames each claim as a natural language inference (NLI) hypoth-  \nesis against retrieved evidence. In biomedicine, however, textual entailment alone is insufficient: a claim can be linguistically supported by retrieved passages yet violate known drug–disease contraindications. Recent work therefore augments NLI with biomedical knowledge graphs (KGs) (Vladika et al., 2024 ; Ning et al., 2025), under the implicit assumption that adding a KG can only help. This assumption is unsafe. Figure 1 illustrates the failure: a claim about ibuprofen’s gastrointestinal side effects receives a spurious CONTRADICT verdict from a KG that lacks the relevant entity pair, yet is correctly entailed by a broader-coverage KG. Whether a KG helps depends on whether it covers the cl","cbCaiueBPXDLDp0h","https://ap.wps.com/l/cbCaiueBPXDLDp0h","pdf",1001615,26,"English","# Abstract\n# 1 Introduction\n## Selective gating in MEDRAGCHECKER\n## Claim-level evaluation with NLI and KG verification\n## Coverage-aware fusion and KG back-ends\n## Contributions","[{\"question\":\"What problem does MEDRAGCHECKER address in biomedical retrieval-augmented generation?\",\"answer\":\"Long-form biomedical RAG answers can contain isolated clinically important errors that whole-answer scoring may hide. MEDRAGCHECKER targets this by verifying claims at a finer granularity.\"},{\"question\":\"How does MEDRAGCHECKER verify biomedical claims?\",\"answer\":\"It decomposes generated answers into atomic claims, then verifies each claim using textual natural language inference (NLI) combined with a knowledge-graph (KG) consistency signal.\"},{\"question\":\"Why does the method use selective gating instead of always trusting the KG?\",\"answer\":\"The KG can help or hurt depending on whether it covers the relevant claim entities and relations. Selective gating fuses the KG signal only when it strongly conflicts with a confident NLI entailment.\"}]","MedRAGChecker - Claim-Level Verification for Biomedical Retrieval-Augmented Generation | PDF",66]