{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:QYNPNYIZNDJOXPTL5SBBCQD5TL","short_pith_number":"pith:QYNPNYIZ","schema_version":"1.0","canonical_sha256":"861af6e11968d2ebbe6bec8211407d9afcaf4eb163a38460f9475bd069fe3f6b","source":{"kind":"arxiv","id":"2406.12702","version":2},"attestation_state":"computed","paper":{"title":"[WIP] Jailbreak Paradox: The Achilles' Heel of LLMs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Abhinav Rao, Monojit Choudhury, Somak Aditya","submitted_at":"2024-06-18T15:14:35Z","abstract_excerpt":"We introduce two paradoxes concerning jailbreak of foundation models: First, it is impossible to construct a perfect jailbreak classifier, and second, a weaker model cannot consistently detect whether a stronger (in a pareto-dominant sense) model is jailbroken or not. We provide formal proofs for these paradoxes and a short case study on Llama and GPT4-o to demonstrate this. We discuss broader theoretical and practical repercussions of these results."},"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":"2406.12702","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-06-18T15:14:35Z","cross_cats_sorted":[],"title_canon_sha256":"176d4e36c52cc1f4cb4dee27fbd5478d9f2eb5d0512fe4b6d4d40f7c984da01c","abstract_canon_sha256":"58a7bcdbba1ec2d48df6220f9f2682f95038f5f4d069fd353a3eb03624135634"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:35:04.393305Z","signature_b64":"TjU61wI6lnTK+s9Gibu1q4XMn+SZGTioRxj2StLq51KlzAf0X/WJoZKm+Gp5c1FsjfiM5BcPBMZEqg2hXgUEAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"861af6e11968d2ebbe6bec8211407d9afcaf4eb163a38460f9475bd069fe3f6b","last_reissued_at":"2026-07-05T08:35:04.392895Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:35:04.392895Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"[WIP] Jailbreak Paradox: The Achilles' Heel of LLMs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Abhinav Rao, Monojit Choudhury, Somak Aditya","submitted_at":"2024-06-18T15:14:35Z","abstract_excerpt":"We introduce two paradoxes concerning jailbreak of foundation models: First, it is impossible to construct a perfect jailbreak classifier, and second, a weaker model cannot consistently detect whether a stronger (in a pareto-dominant sense) model is jailbroken or not. We provide formal proofs for these paradoxes and a short case study on Llama and GPT4-o to demonstrate this. We discuss broader theoretical and practical repercussions of these results."},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.12702","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/2406.12702/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":"2406.12702","created_at":"2026-07-05T08:35:04.392950+00:00"},{"alias_kind":"arxiv_version","alias_value":"2406.12702v2","created_at":"2026-07-05T08:35:04.392950+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.12702","created_at":"2026-07-05T08:35:04.392950+00:00"},{"alias_kind":"pith_short_12","alias_value":"QYNPNYIZNDJO","created_at":"2026-07-05T08:35:04.392950+00:00"},{"alias_kind":"pith_short_16","alias_value":"QYNPNYIZNDJOXPTL","created_at":"2026-07-05T08:35:04.392950+00:00"},{"alias_kind":"pith_short_8","alias_value":"QYNPNYIZ","created_at":"2026-07-05T08:35:04.392950+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2508.07139","citing_title":"A Real-Time, Self-Tuning Moderator Framework for Adversarial Prompt Detection","ref_index":26,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/QYNPNYIZNDJOXPTL5SBBCQD5TL","json":"https://pith.science/pith/QYNPNYIZNDJOXPTL5SBBCQD5TL.json","graph_json":"https://pith.science/api/pith-number/QYNPNYIZNDJOXPTL5SBBCQD5TL/graph.json","events_json":"https://pith.science/api/pith-number/QYNPNYIZNDJOXPTL5SBBCQD5TL/events.json","paper":"https://pith.science/paper/QYNPNYIZ"},"agent_actions":{"view_html":"https://pith.science/pith/QYNPNYIZNDJOXPTL5SBBCQD5TL","download_json":"https://pith.science/pith/QYNPNYIZNDJOXPTL5SBBCQD5TL.json","view_paper":"https://pith.science/paper/QYNPNYIZ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2406.12702&json=true","fetch_graph":"https://pith.science/api/pith-number/QYNPNYIZNDJOXPTL5SBBCQD5TL/graph.json","fetch_events":"https://pith.science/api/pith-number/QYNPNYIZNDJOXPTL5SBBCQD5TL/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/QYNPNYIZNDJOXPTL5SBBCQD5TL/action/timestamp_anchor","attest_storage":"https://pith.science/pith/QYNPNYIZNDJOXPTL5SBBCQD5TL/action/storage_attestation","attest_author":"https://pith.science/pith/QYNPNYIZNDJOXPTL5SBBCQD5TL/action/author_attestation","sign_citation":"https://pith.science/pith/QYNPNYIZNDJOXPTL5SBBCQD5TL/action/citation_signature","submit_replication":"https://pith.science/pith/QYNPNYIZNDJOXPTL5SBBCQD5TL/action/replication_record"}},"created_at":"2026-07-05T08:35:04.392950+00:00","updated_at":"2026-07-05T08:35:04.392950+00:00"}