{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:BF3P6B2J4EQMOXIFWVOS4OY5IA","short_pith_number":"pith:BF3P6B2J","schema_version":"1.0","canonical_sha256":"0976ff0749e120c75d05b55d2e3b1d4033053fa8f39d8b38894f40d4404422ab","source":{"kind":"arxiv","id":"2607.23496","version":1},"attestation_state":"computed","paper":{"title":"Do LLMs Know Their Vulnerable Scenarios?","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CR"],"primary_cat":"cs.AI","authors_text":"Haoran Jing, Huiqi Deng, Jiahui Han, Na Zou, Xia Hu, Xiting Wang, Xuankun Rong, Ziheng Peng","submitted_at":"2026-07-26T06:54:28Z","abstract_excerpt":"Safety-aligned large language models are trained to refuse harmful requests, yet embedding the same requests in particular scenarios can bypass their safeguards. Existing red-teaming methods empirically identify effective scenarios through observed attack outcomes, but why particular scenarios weaken refusal remains mechanistically unclear. Meanwhile, mechanistic interpretability studies have characterized both refusal directions and jailbreak-associated features, without explaining the relationship between the two representations. In this work, we show that scenario-wrapped prompts activate i"},"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":"2607.23496","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2026-07-26T06:54:28Z","cross_cats_sorted":["cs.CR"],"title_canon_sha256":"49f7f9f1fd93b93b053a2f41741a3cb81f86752a57336ecb01feba8d06576ea7","abstract_canon_sha256":"364f4575e39b3576860dcb3df39da74f23edbef7d2d1c84841df2098068b8a47"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-28T01:22:56.048703Z","signature_b64":"RXy7tschq/BkTkOKBKFchY1eL21fehoR1aux7gT4hIGxUxIeDvXO/8r2ouhpuhcjjFDiy/MsJhQfqFo2bJnTCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0976ff0749e120c75d05b55d2e3b1d4033053fa8f39d8b38894f40d4404422ab","last_reissued_at":"2026-07-28T01:22:56.047880Z","signature_status":"signed_v1","first_computed_at":"2026-07-28T01:22:56.047880Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Do LLMs Know Their Vulnerable Scenarios?","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CR"],"primary_cat":"cs.AI","authors_text":"Haoran Jing, Huiqi Deng, Jiahui Han, Na Zou, Xia Hu, Xiting Wang, Xuankun Rong, Ziheng Peng","submitted_at":"2026-07-26T06:54:28Z","abstract_excerpt":"Safety-aligned large language models are trained to refuse harmful requests, yet embedding the same requests in particular scenarios can bypass their safeguards. Existing red-teaming methods empirically identify effective scenarios through observed attack outcomes, but why particular scenarios weaken refusal remains mechanistically unclear. Meanwhile, mechanistic interpretability studies have characterized both refusal directions and jailbreak-associated features, without explaining the relationship between the two representations. In this work, we show that scenario-wrapped prompts activate i"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2607.23496","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/2607.23496/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":"2607.23496","created_at":"2026-07-28T01:22:56.048304+00:00"},{"alias_kind":"arxiv_version","alias_value":"2607.23496v1","created_at":"2026-07-28T01:22:56.048304+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2607.23496","created_at":"2026-07-28T01:22:56.048304+00:00"},{"alias_kind":"pith_short_12","alias_value":"BF3P6B2J4EQM","created_at":"2026-07-28T01:22:56.048304+00:00"},{"alias_kind":"pith_short_16","alias_value":"BF3P6B2J4EQMOXIF","created_at":"2026-07-28T01:22:56.048304+00:00"},{"alias_kind":"pith_short_8","alias_value":"BF3P6B2J","created_at":"2026-07-28T01:22:56.048304+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/BF3P6B2J4EQMOXIFWVOS4OY5IA","json":"https://pith.science/pith/BF3P6B2J4EQMOXIFWVOS4OY5IA.json","graph_json":"https://pith.science/api/pith-number/BF3P6B2J4EQMOXIFWVOS4OY5IA/graph.json","events_json":"https://pith.science/api/pith-number/BF3P6B2J4EQMOXIFWVOS4OY5IA/events.json","paper":"https://pith.science/paper/BF3P6B2J"},"agent_actions":{"view_html":"https://pith.science/pith/BF3P6B2J4EQMOXIFWVOS4OY5IA","download_json":"https://pith.science/pith/BF3P6B2J4EQMOXIFWVOS4OY5IA.json","view_paper":"https://pith.science/paper/BF3P6B2J","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2607.23496&json=true","fetch_graph":"https://pith.science/api/pith-number/BF3P6B2J4EQMOXIFWVOS4OY5IA/graph.json","fetch_events":"https://pith.science/api/pith-number/BF3P6B2J4EQMOXIFWVOS4OY5IA/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/BF3P6B2J4EQMOXIFWVOS4OY5IA/action/timestamp_anchor","attest_storage":"https://pith.science/pith/BF3P6B2J4EQMOXIFWVOS4OY5IA/action/storage_attestation","attest_author":"https://pith.science/pith/BF3P6B2J4EQMOXIFWVOS4OY5IA/action/author_attestation","sign_citation":"https://pith.science/pith/BF3P6B2J4EQMOXIFWVOS4OY5IA/action/citation_signature","submit_replication":"https://pith.science/pith/BF3P6B2J4EQMOXIFWVOS4OY5IA/action/replication_record"}},"created_at":"2026-07-28T01:22:56.048304+00:00","updated_at":"2026-07-28T01:22:56.048304+00:00"}