{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:XBMMC3LRUBL5YOCHWL3HDLDE3Q","short_pith_number":"pith:XBMMC3LR","schema_version":"1.0","canonical_sha256":"b858c16d71a057dc3847b2f671ac64dc10538f17a5df7f9c8560888b6806c9ab","source":{"kind":"arxiv","id":"2405.13077","version":2},"attestation_state":"computed","paper":{"title":"GPT-4 Jailbreaks Itself with Near-Perfect Success Using Self-Explanation","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.CR","authors_text":"Govind Ramesh, Wei Xu, Yao Dou","submitted_at":"2024-05-21T03:16:35Z","abstract_excerpt":"Research on jailbreaking has been valuable for testing and understanding the safety and security issues of large language models (LLMs). In this paper, we introduce Iterative Refinement Induced Self-Jailbreak (IRIS), a novel approach that leverages the reflective capabilities of LLMs for jailbreaking with only black-box access. Unlike previous methods, IRIS simplifies the jailbreaking process by using a single model as both the attacker and target. This method first iteratively refines adversarial prompts through self-explanation, which is crucial for ensuring that even well-aligned LLMs obey "},"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":"2405.13077","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CR","submitted_at":"2024-05-21T03:16:35Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"9de901d27036187cff3e665918f31b731939d94663259768055b5bfb08f73e69","abstract_canon_sha256":"8daa7ae132bb987725a16192dfe862aba75ae032faf8e85e94797e9f6145d638"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:21:04.737866Z","signature_b64":"DWpyXGx6jH0/WAEftCwwXFNA7clRT2H4tns37jRyvbMtD5WhtuAmV4C+yeK7evxNvE9wp746KKXDNfigfWhLAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b858c16d71a057dc3847b2f671ac64dc10538f17a5df7f9c8560888b6806c9ab","last_reissued_at":"2026-07-05T09:21:04.737344Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:21:04.737344Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"GPT-4 Jailbreaks Itself with Near-Perfect Success Using Self-Explanation","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.CR","authors_text":"Govind Ramesh, Wei Xu, Yao Dou","submitted_at":"2024-05-21T03:16:35Z","abstract_excerpt":"Research on jailbreaking has been valuable for testing and understanding the safety and security issues of large language models (LLMs). In this paper, we introduce Iterative Refinement Induced Self-Jailbreak (IRIS), a novel approach that leverages the reflective capabilities of LLMs for jailbreaking with only black-box access. Unlike previous methods, IRIS simplifies the jailbreaking process by using a single model as both the attacker and target. This method first iteratively refines adversarial prompts through self-explanation, which is crucial for ensuring that even well-aligned LLMs obey "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2405.13077","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/2405.13077/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":"2405.13077","created_at":"2026-07-05T09:21:04.737411+00:00"},{"alias_kind":"arxiv_version","alias_value":"2405.13077v2","created_at":"2026-07-05T09:21:04.737411+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2405.13077","created_at":"2026-07-05T09:21:04.737411+00:00"},{"alias_kind":"pith_short_12","alias_value":"XBMMC3LRUBL5","created_at":"2026-07-05T09:21:04.737411+00:00"},{"alias_kind":"pith_short_16","alias_value":"XBMMC3LRUBL5YOCH","created_at":"2026-07-05T09:21:04.737411+00:00"},{"alias_kind":"pith_short_8","alias_value":"XBMMC3LR","created_at":"2026-07-05T09:21:04.737411+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2604.11309","citing_title":"The Salami Slicing Threat: Exploiting Cumulative Risks in LLM Systems","ref_index":30,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/XBMMC3LRUBL5YOCHWL3HDLDE3Q","json":"https://pith.science/pith/XBMMC3LRUBL5YOCHWL3HDLDE3Q.json","graph_json":"https://pith.science/api/pith-number/XBMMC3LRUBL5YOCHWL3HDLDE3Q/graph.json","events_json":"https://pith.science/api/pith-number/XBMMC3LRUBL5YOCHWL3HDLDE3Q/events.json","paper":"https://pith.science/paper/XBMMC3LR"},"agent_actions":{"view_html":"https://pith.science/pith/XBMMC3LRUBL5YOCHWL3HDLDE3Q","download_json":"https://pith.science/pith/XBMMC3LRUBL5YOCHWL3HDLDE3Q.json","view_paper":"https://pith.science/paper/XBMMC3LR","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2405.13077&json=true","fetch_graph":"https://pith.science/api/pith-number/XBMMC3LRUBL5YOCHWL3HDLDE3Q/graph.json","fetch_events":"https://pith.science/api/pith-number/XBMMC3LRUBL5YOCHWL3HDLDE3Q/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/XBMMC3LRUBL5YOCHWL3HDLDE3Q/action/timestamp_anchor","attest_storage":"https://pith.science/pith/XBMMC3LRUBL5YOCHWL3HDLDE3Q/action/storage_attestation","attest_author":"https://pith.science/pith/XBMMC3LRUBL5YOCHWL3HDLDE3Q/action/author_attestation","sign_citation":"https://pith.science/pith/XBMMC3LRUBL5YOCHWL3HDLDE3Q/action/citation_signature","submit_replication":"https://pith.science/pith/XBMMC3LRUBL5YOCHWL3HDLDE3Q/action/replication_record"}},"created_at":"2026-07-05T09:21:04.737411+00:00","updated_at":"2026-07-05T09:21:04.737411+00:00"}