{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:GIDL7EDIPK547X4WFQF343VGDI","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"881f5c372f9beeea9741eb229f4420493808e42c9867d83c4256600ac50ab2cc","cross_cats_sorted":["cs.AI"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-05-10T02:39:05Z","title_canon_sha256":"2157ed097ea7cbccee21b89c015f0df8544df3e69a5bc8306a22aa8959162c6d"},"schema_version":"1.0","source":{"id":"2605.09270","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.09270","created_at":"2026-05-26T02:04:12Z"},{"alias_kind":"arxiv_version","alias_value":"2605.09270v2","created_at":"2026-05-26T02:04:12Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.09270","created_at":"2026-05-26T02:04:12Z"},{"alias_kind":"pith_short_12","alias_value":"GIDL7EDIPK54","created_at":"2026-05-26T02:04:12Z"},{"alias_kind":"pith_short_16","alias_value":"GIDL7EDIPK547X4W","created_at":"2026-05-26T02:04:12Z"},{"alias_kind":"pith_short_8","alias_value":"GIDL7EDI","created_at":"2026-05-26T02:04:12Z"}],"graph_snapshots":[{"event_id":"sha256:bfac1cc55599659f7ffb8c7d5cc1704340b95dfb31467611879584e88c52e93f","target":"graph","created_at":"2026-05-26T02:04:12Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"Generalization failures stem not from memorization as a mechanism, but from memorizing the wrong inductive targets."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That the reported performance gains are caused by the shift to theorem-level supervision rather than by other unspecified differences in data construction, prompting, or training hyperparameters between vanilla SFT and Theorem-SFT."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"Theorem-SFT improves mathematical reasoning generalization by teaching theorem application rather than instance memorization, delivering gains of +8.8% on MATH and +20.27% on GeoQA across model families."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Supervised fine-tuning for math reasoning succeeds when models learn to apply theorems explicitly instead of memorizing individual problem-answer pairs."}],"snapshot_sha256":"2ac3c3605335d47feb2fce4ee18293ae8490f872b3ee468d2f940c37bf81d2ad"},"formal_canon":{"evidence_count":2,"snapshot_sha256":"e90073efb785a850c8bbcb483d17bd3ecc03cc98414f435b4534a6c1a00a2f58"},"integrity":{"available":true,"clean":true,"detectors_run":[{"findings_count":0,"name":"claim_evidence","ran_at":"2026-05-20T08:02:08.944337Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"ai_meta_artifact","ran_at":"2026-05-19T20:34:31.582370Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"doi_title_agreement","ran_at":"2026-05-19T13:31:17.837425Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"doi_compliance","ran_at":"2026-05-19T10:22:17.241840Z","status":"completed","version":"1.0.0"}],"endpoint":"/pith/2605.09270/integrity.json","findings":[],"snapshot_sha256":"83df298cf3e2c2c390afdf01fcaf3a6efeeb248bd79cbd9c70a7c5a47a0dae34","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Supervised Fine-Tuning (SFT) is widely used for task-specific adaptation, yet recent work shows it systematically undermines reasoning generalization. We argue the root cause is not memorization itself, but its target: vanilla SFT drives models to exploit and memorize spurious surface correlations in problem-solution pairs, leaving them brittle to superficial input variations. To address this, we propose Theorem-SFT, which reorients supervision toward explicit theorem application by teaching models how rules are invoked rather than what answers look like. Theorem-SFT yields consistent gains ac","authors_text":"Jing Lei, Mengyu Yang, Ruiying Peng, Xiaohui Li, Xinlei Chen, Xueyu Wu","cross_cats":["cs.AI"],"headline":"Supervised fine-tuning for math reasoning succeeds when models learn to apply theorems explicitly instead of memorizing individual problem-answer pairs.","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-05-10T02:39:05Z","title":"Memorize Theorems, Not Instances: Probing SFT Generalization through Mathematical Reasoning"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.09270","kind":"arxiv","version":2},"verdict":{"created_at":"2026-05-12T04:47:26.502142Z","id":"d59a00fe-0325-4e05-a41f-c3b14d40b8ff","model_set":{"reader":"grok-4.3"},"one_line_summary":"Theorem-SFT improves mathematical reasoning generalization by teaching theorem application rather than instance memorization, delivering gains of +8.8% on MATH and +20.27% on GeoQA across model families.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Supervised fine-tuning for math reasoning succeeds when models learn to apply theorems explicitly instead of memorizing individual problem-answer pairs.","strongest_claim":"Generalization failures stem not from memorization as a mechanism, but from memorizing the wrong inductive targets.","weakest_assumption":"That the reported performance gains are caused by the shift to theorem-level supervision rather than by other unspecified differences in data construction, prompting, or training hyperparameters between vanilla SFT and Theorem-SFT."}},"verdict_id":"d59a00fe-0325-4e05-a41f-c3b14d40b8ff"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:1c2e487db8914692e5f53f1aec527036799749bfd66f88aad15a7eb3d4aae86e","target":"record","created_at":"2026-05-26T02:04:12Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"881f5c372f9beeea9741eb229f4420493808e42c9867d83c4256600ac50ab2cc","cross_cats_sorted":["cs.AI"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-05-10T02:39:05Z","title_canon_sha256":"2157ed097ea7cbccee21b89c015f0df8544df3e69a5bc8306a22aa8959162c6d"},"schema_version":"1.0","source":{"id":"2605.09270","kind":"arxiv","version":2}},"canonical_sha256":"3206bf90687abbcfdf962c0bbe6ea61a259ffc8a711e099e8cd6a8c01927ea9e","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"3206bf90687abbcfdf962c0bbe6ea61a259ffc8a711e099e8cd6a8c01927ea9e","first_computed_at":"2026-05-26T02:04:12.658778Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-26T02:04:12.658778Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"i24naM5FCvKmx0u21Rv+fuhGJB7oRATMLbjtHsCOBm4QtVwdPPWznmonmuYdSRcTg7akYsZqbIKpmp+Eht2aDw==","signature_status":"signed_v1","signed_at":"2026-05-26T02:04:12.659552Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.09270","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:1c2e487db8914692e5f53f1aec527036799749bfd66f88aad15a7eb3d4aae86e","sha256:bfac1cc55599659f7ffb8c7d5cc1704340b95dfb31467611879584e88c52e93f"],"state_sha256":"f2885019e2e3e57ddc9c3e68c3ad8bfc0259187f7434190224d8a34956d0ebd7"}