{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:7XCHDNJ7ONAYEURW5GM5IXKX3J","short_pith_number":"pith:7XCHDNJ7","schema_version":"1.0","canonical_sha256":"fdc471b53f7341825236e999d45d57da7cd514dd2259e062142897cc216f0888","source":{"kind":"arxiv","id":"2605.28301","version":1},"attestation_state":"computed","paper":{"title":"Better Accuracies, Worse Reasoning: A Step-Level Audit of Medical Chain-of-Thought Distillation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Fei Teng, Honghan Wu, Jiacong Mi, Xuanqi Peng, Yunsoo Kim, Zhaoyang Jiang, Zhizhong Fu, Zicheng Li","submitted_at":"2026-05-27T10:55:24Z","abstract_excerpt":"Chain-of-thought (CoT) distillation trains a smaller model to imitate a teacher's reasoning trace, but it is typically evaluated by final-answer metrics including accuracy. We ask whether gains in answer quality are accompanied by improvements in the trace. In medical QA, where short answer options can leave a richer clinical justification under-specified, a Qwen3-8B student distilled from a DeepSeek-V3-family teacher improves on MedQA-USMLE answer metrics (SC@64 74.7% to 84.4%; expected calibration error (ECE) 0.096 to 0.034). Yet under a Kimi-K2.6 style-blind LLM-judge audit, its error rate "},"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":"2605.28301","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2026-05-27T10:55:24Z","cross_cats_sorted":[],"title_canon_sha256":"349ac5e884accf1c74cc477ce8432ffb4fa996d35db9a38348d640a5007d4bac","abstract_canon_sha256":"6974a9e955d8949c29a990d9ad52ae5b595de030971357baba7d6c6abc88831d"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-28T01:05:05.713678Z","signature_b64":"Pr9hK5HPxUJKQ3gL1s1vS6xY2MAaz0AbwBY0Er5dPLmLVozooOiEGMuyJ5qMIZso+KYcOtE6UI1uecVyf4I0DQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"fdc471b53f7341825236e999d45d57da7cd514dd2259e062142897cc216f0888","last_reissued_at":"2026-05-28T01:05:05.713253Z","signature_status":"signed_v1","first_computed_at":"2026-05-28T01:05:05.713253Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Better Accuracies, Worse Reasoning: A Step-Level Audit of Medical Chain-of-Thought Distillation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Fei Teng, Honghan Wu, Jiacong Mi, Xuanqi Peng, Yunsoo Kim, Zhaoyang Jiang, Zhizhong Fu, Zicheng Li","submitted_at":"2026-05-27T10:55:24Z","abstract_excerpt":"Chain-of-thought (CoT) distillation trains a smaller model to imitate a teacher's reasoning trace, but it is typically evaluated by final-answer metrics including accuracy. We ask whether gains in answer quality are accompanied by improvements in the trace. In medical QA, where short answer options can leave a richer clinical justification under-specified, a Qwen3-8B student distilled from a DeepSeek-V3-family teacher improves on MedQA-USMLE answer metrics (SC@64 74.7% to 84.4%; expected calibration error (ECE) 0.096 to 0.034). Yet under a Kimi-K2.6 style-blind LLM-judge audit, its error rate "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.28301","kind":"arxiv","version":1},"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/2605.28301/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":"2605.28301","created_at":"2026-05-28T01:05:05.713311+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.28301v1","created_at":"2026-05-28T01:05:05.713311+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.28301","created_at":"2026-05-28T01:05:05.713311+00:00"},{"alias_kind":"pith_short_12","alias_value":"7XCHDNJ7ONAY","created_at":"2026-05-28T01:05:05.713311+00:00"},{"alias_kind":"pith_short_16","alias_value":"7XCHDNJ7ONAYEURW","created_at":"2026-05-28T01:05:05.713311+00:00"},{"alias_kind":"pith_short_8","alias_value":"7XCHDNJ7","created_at":"2026-05-28T01:05:05.713311+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/7XCHDNJ7ONAYEURW5GM5IXKX3J","json":"https://pith.science/pith/7XCHDNJ7ONAYEURW5GM5IXKX3J.json","graph_json":"https://pith.science/api/pith-number/7XCHDNJ7ONAYEURW5GM5IXKX3J/graph.json","events_json":"https://pith.science/api/pith-number/7XCHDNJ7ONAYEURW5GM5IXKX3J/events.json","paper":"https://pith.science/paper/7XCHDNJ7"},"agent_actions":{"view_html":"https://pith.science/pith/7XCHDNJ7ONAYEURW5GM5IXKX3J","download_json":"https://pith.science/pith/7XCHDNJ7ONAYEURW5GM5IXKX3J.json","view_paper":"https://pith.science/paper/7XCHDNJ7","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.28301&json=true","fetch_graph":"https://pith.science/api/pith-number/7XCHDNJ7ONAYEURW5GM5IXKX3J/graph.json","fetch_events":"https://pith.science/api/pith-number/7XCHDNJ7ONAYEURW5GM5IXKX3J/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/7XCHDNJ7ONAYEURW5GM5IXKX3J/action/timestamp_anchor","attest_storage":"https://pith.science/pith/7XCHDNJ7ONAYEURW5GM5IXKX3J/action/storage_attestation","attest_author":"https://pith.science/pith/7XCHDNJ7ONAYEURW5GM5IXKX3J/action/author_attestation","sign_citation":"https://pith.science/pith/7XCHDNJ7ONAYEURW5GM5IXKX3J/action/citation_signature","submit_replication":"https://pith.science/pith/7XCHDNJ7ONAYEURW5GM5IXKX3J/action/replication_record"}},"created_at":"2026-05-28T01:05:05.713311+00:00","updated_at":"2026-05-28T01:05:05.713311+00:00"}