{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:AOXK7XWTBWD5RUHWCTU7BYALRJ","short_pith_number":"pith:AOXK7XWT","schema_version":"1.0","canonical_sha256":"03aeafded30d87d8d0f614e9f0e00b8a7ad3ae6e12c1fba31b19dda08dc52d3a","source":{"kind":"arxiv","id":"2505.15214","version":2},"attestation_state":"computed","paper":{"title":"R-TOFU: Unlearning in Large Reasoning Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Albert No, Sangyeon Yoon, Wonje Jeung","submitted_at":"2025-05-21T07:44:30Z","abstract_excerpt":"Large Reasoning Models (LRMs) embed private or copyrighted information not only in their final answers but also throughout multi-step chain-of-thought (CoT) traces, making reliable unlearning far more demanding than in standard LLMs. We introduce Reasoning-TOFU (R-TOFU), the first benchmark tailored to this setting. R-TOFU augments existing unlearning tasks with realistic CoT annotations and provides step-wise metrics that expose residual knowledge invisible to answer-level checks. Using R-TOFU, we carry out a comprehensive comparison of gradient-based and preference-optimization baselines and"},"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":"2505.15214","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-05-21T07:44:30Z","cross_cats_sorted":[],"title_canon_sha256":"cb5edf788c295667996a97470755fad859509bd0176c9f21bac2b47fd5817ef0","abstract_canon_sha256":"9255f77ad14ea4cb5f1f09680b3d89b0b4965416729f53e4e85540f68384b5af"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:10:09.000820Z","signature_b64":"p2etNeGZftcrTSzL0XO2qHRt41JytcYAtYtFQGXJz5cW65cncU5fWRnjXgIAZOYdEXVkcwo13o07gGqpW2IPCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"03aeafded30d87d8d0f614e9f0e00b8a7ad3ae6e12c1fba31b19dda08dc52d3a","last_reissued_at":"2026-07-05T11:10:09.000288Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:10:09.000288Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"R-TOFU: Unlearning in Large Reasoning Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Albert No, Sangyeon Yoon, Wonje Jeung","submitted_at":"2025-05-21T07:44:30Z","abstract_excerpt":"Large Reasoning Models (LRMs) embed private or copyrighted information not only in their final answers but also throughout multi-step chain-of-thought (CoT) traces, making reliable unlearning far more demanding than in standard LLMs. We introduce Reasoning-TOFU (R-TOFU), the first benchmark tailored to this setting. R-TOFU augments existing unlearning tasks with realistic CoT annotations and provides step-wise metrics that expose residual knowledge invisible to answer-level checks. Using R-TOFU, we carry out a comprehensive comparison of gradient-based and preference-optimization baselines and"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.15214","kind":"arxiv","version":2},"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/2505.15214/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":"2505.15214","created_at":"2026-07-05T11:10:09.000369+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.15214v2","created_at":"2026-07-05T11:10:09.000369+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.15214","created_at":"2026-07-05T11:10:09.000369+00:00"},{"alias_kind":"pith_short_12","alias_value":"AOXK7XWTBWD5","created_at":"2026-07-05T11:10:09.000369+00:00"},{"alias_kind":"pith_short_16","alias_value":"AOXK7XWTBWD5RUHW","created_at":"2026-07-05T11:10:09.000369+00:00"},{"alias_kind":"pith_short_8","alias_value":"AOXK7XWT","created_at":"2026-07-05T11:10:09.000369+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2604.18946","citing_title":"Reasoning Structure Matters for Safety Alignment of Reasoning Models","ref_index":48,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/AOXK7XWTBWD5RUHWCTU7BYALRJ","json":"https://pith.science/pith/AOXK7XWTBWD5RUHWCTU7BYALRJ.json","graph_json":"https://pith.science/api/pith-number/AOXK7XWTBWD5RUHWCTU7BYALRJ/graph.json","events_json":"https://pith.science/api/pith-number/AOXK7XWTBWD5RUHWCTU7BYALRJ/events.json","paper":"https://pith.science/paper/AOXK7XWT"},"agent_actions":{"view_html":"https://pith.science/pith/AOXK7XWTBWD5RUHWCTU7BYALRJ","download_json":"https://pith.science/pith/AOXK7XWTBWD5RUHWCTU7BYALRJ.json","view_paper":"https://pith.science/paper/AOXK7XWT","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.15214&json=true","fetch_graph":"https://pith.science/api/pith-number/AOXK7XWTBWD5RUHWCTU7BYALRJ/graph.json","fetch_events":"https://pith.science/api/pith-number/AOXK7XWTBWD5RUHWCTU7BYALRJ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/AOXK7XWTBWD5RUHWCTU7BYALRJ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/AOXK7XWTBWD5RUHWCTU7BYALRJ/action/storage_attestation","attest_author":"https://pith.science/pith/AOXK7XWTBWD5RUHWCTU7BYALRJ/action/author_attestation","sign_citation":"https://pith.science/pith/AOXK7XWTBWD5RUHWCTU7BYALRJ/action/citation_signature","submit_replication":"https://pith.science/pith/AOXK7XWTBWD5RUHWCTU7BYALRJ/action/replication_record"}},"created_at":"2026-07-05T11:10:09.000369+00:00","updated_at":"2026-07-05T11:10:09.000369+00:00"}