{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:P6SH75ZDTSN2K7BGOPU3OSDVVE","short_pith_number":"pith:P6SH75ZD","schema_version":"1.0","canonical_sha256":"7fa47ff7239c9ba57c2673e9b74875a932d70e572f3d45c1071ec86ada4271ab","source":{"kind":"arxiv","id":"2502.19041","version":2},"attestation_state":"computed","paper":{"title":"Beyond Surface-Level Patterns: An Essence-Driven Defense Framework Against Jailbreak Attacks in LLMs","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CR","authors_text":"Ansen Zhang, Ronghao Chen, Shiyu Xiang, Yanfei Cao, Yang Fan","submitted_at":"2025-02-26T10:53:58Z","abstract_excerpt":"Although Aligned Large Language Models (LLMs) are trained to refuse harmful requests, they remain vulnerable to jailbreak attacks. Unfortunately, existing methods often focus on surface-level patterns, overlooking the deeper attack essences. As a result, defenses fail when attack prompts change, even though the underlying \"attack essence\" remains the same. To address this issue, we introduce EDDF, an \\textbf{E}ssence-\\textbf{D}riven \\textbf{D}efense \\textbf{F}ramework Against Jailbreak Attacks in LLMs. EDDF is a plug-and-play input-filtering method and operates in two stages: 1) offline essenc"},"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":"2502.19041","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CR","submitted_at":"2025-02-26T10:53:58Z","cross_cats_sorted":[],"title_canon_sha256":"7cff4d4ea1f123f56c3d67ffb1e9338544199fc72ad6bbde26134c1942674af1","abstract_canon_sha256":"ae883960b29417f2081a9e3376721e940205641a6089480cc7c422197f39316e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:10:53.465731Z","signature_b64":"bBHMeOg1WPb9G+c/VaA9GKcoavc/k02RzSYUEkN9GgAlOC3tdoQ48awJtXrjv5S0/JKKsdsxFtxdImK3JJIMDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"7fa47ff7239c9ba57c2673e9b74875a932d70e572f3d45c1071ec86ada4271ab","last_reissued_at":"2026-07-05T11:10:53.465169Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:10:53.465169Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Beyond Surface-Level Patterns: An Essence-Driven Defense Framework Against Jailbreak Attacks in LLMs","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CR","authors_text":"Ansen Zhang, Ronghao Chen, Shiyu Xiang, Yanfei Cao, Yang Fan","submitted_at":"2025-02-26T10:53:58Z","abstract_excerpt":"Although Aligned Large Language Models (LLMs) are trained to refuse harmful requests, they remain vulnerable to jailbreak attacks. Unfortunately, existing methods often focus on surface-level patterns, overlooking the deeper attack essences. As a result, defenses fail when attack prompts change, even though the underlying \"attack essence\" remains the same. To address this issue, we introduce EDDF, an \\textbf{E}ssence-\\textbf{D}riven \\textbf{D}efense \\textbf{F}ramework Against Jailbreak Attacks in LLMs. EDDF is a plug-and-play input-filtering method and operates in two stages: 1) offline essenc"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.19041","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/2502.19041/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":"2502.19041","created_at":"2026-07-05T11:10:53.465236+00:00"},{"alias_kind":"arxiv_version","alias_value":"2502.19041v2","created_at":"2026-07-05T11:10:53.465236+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.19041","created_at":"2026-07-05T11:10:53.465236+00:00"},{"alias_kind":"pith_short_12","alias_value":"P6SH75ZDTSN2","created_at":"2026-07-05T11:10:53.465236+00:00"},{"alias_kind":"pith_short_16","alias_value":"P6SH75ZDTSN2K7BG","created_at":"2026-07-05T11:10:53.465236+00:00"},{"alias_kind":"pith_short_8","alias_value":"P6SH75ZD","created_at":"2026-07-05T11:10:53.465236+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2510.20129","citing_title":"SAID: Safety-Aware Intent Defense via Prefix Probing for Large Language Models","ref_index":20,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/P6SH75ZDTSN2K7BGOPU3OSDVVE","json":"https://pith.science/pith/P6SH75ZDTSN2K7BGOPU3OSDVVE.json","graph_json":"https://pith.science/api/pith-number/P6SH75ZDTSN2K7BGOPU3OSDVVE/graph.json","events_json":"https://pith.science/api/pith-number/P6SH75ZDTSN2K7BGOPU3OSDVVE/events.json","paper":"https://pith.science/paper/P6SH75ZD"},"agent_actions":{"view_html":"https://pith.science/pith/P6SH75ZDTSN2K7BGOPU3OSDVVE","download_json":"https://pith.science/pith/P6SH75ZDTSN2K7BGOPU3OSDVVE.json","view_paper":"https://pith.science/paper/P6SH75ZD","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2502.19041&json=true","fetch_graph":"https://pith.science/api/pith-number/P6SH75ZDTSN2K7BGOPU3OSDVVE/graph.json","fetch_events":"https://pith.science/api/pith-number/P6SH75ZDTSN2K7BGOPU3OSDVVE/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/P6SH75ZDTSN2K7BGOPU3OSDVVE/action/timestamp_anchor","attest_storage":"https://pith.science/pith/P6SH75ZDTSN2K7BGOPU3OSDVVE/action/storage_attestation","attest_author":"https://pith.science/pith/P6SH75ZDTSN2K7BGOPU3OSDVVE/action/author_attestation","sign_citation":"https://pith.science/pith/P6SH75ZDTSN2K7BGOPU3OSDVVE/action/citation_signature","submit_replication":"https://pith.science/pith/P6SH75ZDTSN2K7BGOPU3OSDVVE/action/replication_record"}},"created_at":"2026-07-05T11:10:53.465236+00:00","updated_at":"2026-07-05T11:10:53.465236+00:00"}