{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:CINQIN2WTR3MX47L6V4FHLLJ3I","short_pith_number":"pith:CINQIN2W","schema_version":"1.0","canonical_sha256":"121b0437569c76cbf3ebf57853ad69da3ae63147a4665a6e22becc154901a1a3","source":{"kind":"arxiv","id":"2505.17650","version":1},"attestation_state":"computed","paper":{"title":"Does Chain-of-Thought Reasoning Really Reduce Harmfulness from Jailbreaking?","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Chengda Lu, Jijie Li, Rongwu Xu, Wei Xu, Xiaoyu Fan, Yu Huang","submitted_at":"2025-05-23T09:14:48Z","abstract_excerpt":"Jailbreak attacks have been observed to largely fail against recent reasoning models enhanced by Chain-of-Thought (CoT) reasoning. However, the underlying mechanism remains underexplored, and relying solely on reasoning capacity may raise security concerns. In this paper, we try to answer the question: Does CoT reasoning really reduce harmfulness from jailbreaking? Through rigorous theoretical analysis, we demonstrate that CoT reasoning has dual effects on jailbreaking harmfulness. Based on the theoretical insights, we propose a novel jailbreak method, FicDetail, whose practical performance va"},"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":"2505.17650","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2025-05-23T09:14:48Z","cross_cats_sorted":[],"title_canon_sha256":"749199ca29160c54db3ebddea655fd5516c7b82047b9ec3c8d2df3cf5ed1d837","abstract_canon_sha256":"d447b69b5ed8075cfc14d32b0f0b4c2089e3078b0443aeb4d4c99b5ec82a9a07"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:08:29.905942Z","signature_b64":"pQW3ygjvIhbwAp0mCY/D0qXzEG962XJEZYOGUxVRx+OVcitRnowXdmI/abM+RqZaZfb4KStuVSCPXuo5iPqNCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"121b0437569c76cbf3ebf57853ad69da3ae63147a4665a6e22becc154901a1a3","last_reissued_at":"2026-07-05T11:08:29.905494Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:08:29.905494Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Does Chain-of-Thought Reasoning Really Reduce Harmfulness from Jailbreaking?","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Chengda Lu, Jijie Li, Rongwu Xu, Wei Xu, Xiaoyu Fan, Yu Huang","submitted_at":"2025-05-23T09:14:48Z","abstract_excerpt":"Jailbreak attacks have been observed to largely fail against recent reasoning models enhanced by Chain-of-Thought (CoT) reasoning. However, the underlying mechanism remains underexplored, and relying solely on reasoning capacity may raise security concerns. In this paper, we try to answer the question: Does CoT reasoning really reduce harmfulness from jailbreaking? Through rigorous theoretical analysis, we demonstrate that CoT reasoning has dual effects on jailbreaking harmfulness. Based on the theoretical insights, we propose a novel jailbreak method, FicDetail, whose practical performance va"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.17650","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/2505.17650/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":"2505.17650","created_at":"2026-07-05T11:08:29.905553+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.17650v1","created_at":"2026-07-05T11:08:29.905553+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.17650","created_at":"2026-07-05T11:08:29.905553+00:00"},{"alias_kind":"pith_short_12","alias_value":"CINQIN2WTR3M","created_at":"2026-07-05T11:08:29.905553+00:00"},{"alias_kind":"pith_short_16","alias_value":"CINQIN2WTR3MX47L","created_at":"2026-07-05T11:08:29.905553+00:00"},{"alias_kind":"pith_short_8","alias_value":"CINQIN2W","created_at":"2026-07-05T11:08:29.905553+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/CINQIN2WTR3MX47L6V4FHLLJ3I","json":"https://pith.science/pith/CINQIN2WTR3MX47L6V4FHLLJ3I.json","graph_json":"https://pith.science/api/pith-number/CINQIN2WTR3MX47L6V4FHLLJ3I/graph.json","events_json":"https://pith.science/api/pith-number/CINQIN2WTR3MX47L6V4FHLLJ3I/events.json","paper":"https://pith.science/paper/CINQIN2W"},"agent_actions":{"view_html":"https://pith.science/pith/CINQIN2WTR3MX47L6V4FHLLJ3I","download_json":"https://pith.science/pith/CINQIN2WTR3MX47L6V4FHLLJ3I.json","view_paper":"https://pith.science/paper/CINQIN2W","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.17650&json=true","fetch_graph":"https://pith.science/api/pith-number/CINQIN2WTR3MX47L6V4FHLLJ3I/graph.json","fetch_events":"https://pith.science/api/pith-number/CINQIN2WTR3MX47L6V4FHLLJ3I/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/CINQIN2WTR3MX47L6V4FHLLJ3I/action/timestamp_anchor","attest_storage":"https://pith.science/pith/CINQIN2WTR3MX47L6V4FHLLJ3I/action/storage_attestation","attest_author":"https://pith.science/pith/CINQIN2WTR3MX47L6V4FHLLJ3I/action/author_attestation","sign_citation":"https://pith.science/pith/CINQIN2WTR3MX47L6V4FHLLJ3I/action/citation_signature","submit_replication":"https://pith.science/pith/CINQIN2WTR3MX47L6V4FHLLJ3I/action/replication_record"}},"created_at":"2026-07-05T11:08:29.905553+00:00","updated_at":"2026-07-05T11:08:29.905553+00:00"}