{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:TAXHFACXOXZHYT3TLPELXIVANA","short_pith_number":"pith:TAXHFACX","schema_version":"1.0","canonical_sha256":"982e72805775f27c4f735bc8bba2a0680b5fbe5c1b82f2225a3f49d7c3008865","source":{"kind":"arxiv","id":"2503.07459","version":3},"attestation_state":"computed","paper":{"title":"MedicalAgentsBench for Complex Medical Reasoning: Comparing Internalized Reasoning Models versus Externalized Agent-based Frameworks","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Arman Cohan, Chenglin Wu, Fang Wu, Jiapeng Chen, Jiayi Zhang, Jinyu Xiang, Jiwoong Sohn, Mark Gerstein, Wenqi Shi, Xiangru Tang, Yanjun Shao, Yilun Zhao, Yuxuan Liao","submitted_at":"2025-03-10T15:38:44Z","abstract_excerpt":"Complex medical reasoning requires integrating heterogeneous clinical evidence across multiple inference steps. Large language models (LLMs) now approach this through two routes: internalized reasoning and externalized agent scaffolding (frameworks that decompose problems collaboratively amongst multiple LLMs). To determine whether these routes are exclusive or complementary, we introduce MedicalAgentsBench, a filtered benchmark of 862 complex clinical questions drawn from the union of eight medical datasets via difficulty-aware curation and contamination screening. Evaluating three internaliz"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2503.07459","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-03-10T15:38:44Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"65078ee46de330ed2af33f2f61d41efc29fe3aa4720b77d17b503f6a09e09200","abstract_canon_sha256":"01b4c2d32c3c818b9760876afea3d082116af6b94b7b44402b487b689324fa6b"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-19T16:10:29.419800Z","signature_b64":"rAb8yceQXEuD0nfIoJ6cvTSIcL9Zktb1Koak0b5tvN989E0VNusWhdOAYl8xIlcWU1mXNI08DGnPnoT8acWNAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"982e72805775f27c4f735bc8bba2a0680b5fbe5c1b82f2225a3f49d7c3008865","last_reissued_at":"2026-06-19T16:10:29.419374Z","signature_status":"signed_v1","first_computed_at":"2026-06-19T16:10:29.419374Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"MedicalAgentsBench for Complex Medical Reasoning: Comparing Internalized Reasoning Models versus Externalized Agent-based Frameworks","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Arman Cohan, Chenglin Wu, Fang Wu, Jiapeng Chen, Jiayi Zhang, Jinyu Xiang, Jiwoong Sohn, Mark Gerstein, Wenqi Shi, Xiangru Tang, Yanjun Shao, Yilun Zhao, Yuxuan Liao","submitted_at":"2025-03-10T15:38:44Z","abstract_excerpt":"Complex medical reasoning requires integrating heterogeneous clinical evidence across multiple inference steps. Large language models (LLMs) now approach this through two routes: internalized reasoning and externalized agent scaffolding (frameworks that decompose problems collaboratively amongst multiple LLMs). To determine whether these routes are exclusive or complementary, we introduce MedicalAgentsBench, a filtered benchmark of 862 complex clinical questions drawn from the union of eight medical datasets via difficulty-aware curation and contamination screening. Evaluating three internaliz"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2503.07459","kind":"arxiv","version":3},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2503.07459/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"aliases":[{"alias_kind":"arxiv","alias_value":"2503.07459","created_at":"2026-06-19T16:10:29.419433+00:00"},{"alias_kind":"arxiv_version","alias_value":"2503.07459v3","created_at":"2026-06-19T16:10:29.419433+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2503.07459","created_at":"2026-06-19T16:10:29.419433+00:00"},{"alias_kind":"pith_short_12","alias_value":"TAXHFACXOXZH","created_at":"2026-06-19T16:10:29.419433+00:00"},{"alias_kind":"pith_short_16","alias_value":"TAXHFACXOXZHYT3T","created_at":"2026-06-19T16:10:29.419433+00:00"},{"alias_kind":"pith_short_8","alias_value":"TAXHFACX","created_at":"2026-06-19T16:10:29.419433+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":10,"internal_anchor_count":10,"sample":[{"citing_arxiv_id":"2605.16679","citing_title":"CHI-Bench: Can AI Agents Automate End-to-End, Long-Horizon, Policy-Rich Healthcare Workflows?","ref_index":53,"is_internal_anchor":true},{"citing_arxiv_id":"2605.18652","citing_title":"MementoGUI: Learning Agentic Multimodal Memory Control for Long-Horizon GUI Agents","ref_index":54,"is_internal_anchor":true},{"citing_arxiv_id":"2509.22258","citing_title":"Beyond Classification Accuracy: Neural-MedBench and the Need for Deeper Reasoning Benchmarks","ref_index":27,"is_internal_anchor":true},{"citing_arxiv_id":"2602.12705","citing_title":"MedXIAOHE: A Comprehensive Recipe for Building Medical MLLMs","ref_index":57,"is_internal_anchor":true},{"citing_arxiv_id":"2503.12605","citing_title":"Multimodal Chain-of-Thought Reasoning: A Comprehensive Survey","ref_index":262,"is_internal_anchor":true},{"citing_arxiv_id":"2504.19678","citing_title":"From LLM Reasoning to Autonomous AI Agents: A Comprehensive Review","ref_index":107,"is_internal_anchor":true},{"citing_arxiv_id":"2502.17419","citing_title":"From System 1 to System 2: A Survey of Reasoning Large Language Models","ref_index":119,"is_internal_anchor":true},{"citing_arxiv_id":"2604.24186","citing_title":"MultiDx: A Multi-Source Knowledge Integration Framework towards Diagnostic Reasoning","ref_index":4,"is_internal_anchor":true},{"citing_arxiv_id":"2604.06262","citing_title":"From Exposure to Internalization: Dual-Stream Calibration for In-context Clinical Reasoning","ref_index":59,"is_internal_anchor":true},{"citing_arxiv_id":"2605.01417","citing_title":"Medmarks: A Comprehensive Open-Source LLM Benchmark Suite for Medical Tasks","ref_index":4,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/TAXHFACXOXZHYT3TLPELXIVANA","json":"https://pith.science/pith/TAXHFACXOXZHYT3TLPELXIVANA.json","graph_json":"https://pith.science/api/pith-number/TAXHFACXOXZHYT3TLPELXIVANA/graph.json","events_json":"https://pith.science/api/pith-number/TAXHFACXOXZHYT3TLPELXIVANA/events.json","paper":"https://pith.science/paper/TAXHFACX"},"agent_actions":{"view_html":"https://pith.science/pith/TAXHFACXOXZHYT3TLPELXIVANA","download_json":"https://pith.science/pith/TAXHFACXOXZHYT3TLPELXIVANA.json","view_paper":"https://pith.science/paper/TAXHFACX","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2503.07459&json=true","fetch_graph":"https://pith.science/api/pith-number/TAXHFACXOXZHYT3TLPELXIVANA/graph.json","fetch_events":"https://pith.science/api/pith-number/TAXHFACXOXZHYT3TLPELXIVANA/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/TAXHFACXOXZHYT3TLPELXIVANA/action/timestamp_anchor","attest_storage":"https://pith.science/pith/TAXHFACXOXZHYT3TLPELXIVANA/action/storage_attestation","attest_author":"https://pith.science/pith/TAXHFACXOXZHYT3TLPELXIVANA/action/author_attestation","sign_citation":"https://pith.science/pith/TAXHFACXOXZHYT3TLPELXIVANA/action/citation_signature","submit_replication":"https://pith.science/pith/TAXHFACXOXZHYT3TLPELXIVANA/action/replication_record"}},"created_at":"2026-06-19T16:10:29.419433+00:00","updated_at":"2026-06-19T16:10:29.419433+00:00"}