{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2024:7KFCMEZLG2AU6FGYFTHZQMEQ6E","short_pith_number":"pith:7KFCMEZL","canonical_record":{"source":{"id":"2407.11969","kind":"arxiv","version":4},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-07-16T17:59:55Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"2e07a7126588b9b4682dc080eeaed96734570bc22985626ede93f0185a804058","abstract_canon_sha256":"3e60ea14c28940bd3776853c95e01b6bdfcafb8750549fd3033a02895eb20075"},"schema_version":"1.0"},"canonical_sha256":"fa8a26132b36814f14d82ccf983090f1387bcb8ffd5e4a713c979345aed5160a","source":{"kind":"arxiv","id":"2407.11969","version":4},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2407.11969","created_at":"2026-07-05T10:50:42Z"},{"alias_kind":"arxiv_version","alias_value":"2407.11969v4","created_at":"2026-07-05T10:50:42Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2407.11969","created_at":"2026-07-05T10:50:42Z"},{"alias_kind":"pith_short_12","alias_value":"7KFCMEZLG2AU","created_at":"2026-07-05T10:50:42Z"},{"alias_kind":"pith_short_16","alias_value":"7KFCMEZLG2AU6FGY","created_at":"2026-07-05T10:50:42Z"},{"alias_kind":"pith_short_8","alias_value":"7KFCMEZL","created_at":"2026-07-05T10:50:42Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2024:7KFCMEZLG2AU6FGYFTHZQMEQ6E","target":"record","payload":{"canonical_record":{"source":{"id":"2407.11969","kind":"arxiv","version":4},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-07-16T17:59:55Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"2e07a7126588b9b4682dc080eeaed96734570bc22985626ede93f0185a804058","abstract_canon_sha256":"3e60ea14c28940bd3776853c95e01b6bdfcafb8750549fd3033a02895eb20075"},"schema_version":"1.0"},"canonical_sha256":"fa8a26132b36814f14d82ccf983090f1387bcb8ffd5e4a713c979345aed5160a","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:50:42.939304Z","signature_b64":"eoQ/4f7tBxbXjuTmHg8GYsaHyhlrT4hCodIUADWk4Yj8usNdw+l9ArqKC0MITQAlfVEZXF9CT91aHxNkBfmYCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"fa8a26132b36814f14d82ccf983090f1387bcb8ffd5e4a713c979345aed5160a","last_reissued_at":"2026-07-05T10:50:42.938798Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:50:42.938798Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2407.11969","source_version":4,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T10:50:42Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"Heueij/3N6gyr3l6Cd1Cr21ut+5kHu7goCXbTdVFtwOiNBz1LEy2coAFgYyvw7vSipcXfj/F1ywi8bULUrWkBA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-07-07T11:17:15.808242Z"},"content_sha256":"838e68b9a18fcdd632cd4856245f480b605d4d6a85608dd37b845d714b744f5e","schema_version":"1.0","event_id":"sha256:838e68b9a18fcdd632cd4856245f480b605d4d6a85608dd37b845d714b744f5e"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2024:7KFCMEZLG2AU6FGYFTHZQMEQ6E","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Does Refusal Training in LLMs Generalize to the Past Tense?","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Maksym Andriushchenko, Nicolas Flammarion","submitted_at":"2024-07-16T17:59:55Z","abstract_excerpt":"Refusal training is widely used to prevent LLMs from generating harmful, undesirable, or illegal outputs. We reveal a curious generalization gap in the current refusal training approaches: simply reformulating a harmful request in the past tense (e.g., \"How to make a Molotov cocktail?\" to \"How did people make a Molotov cocktail?\") is often sufficient to jailbreak many state-of-the-art LLMs. We systematically evaluate this method on Llama-3 8B, Claude-3.5 Sonnet, GPT-3.5 Turbo, Gemma-2 9B, Phi-3-Mini, GPT-4o mini, GPT-4o, o1-mini, o1-preview, and R2D2 models using GPT-3.5 Turbo as a reformulati"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2407.11969","kind":"arxiv","version":4},"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/2407.11969/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"},"verdict_id":null},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T10:50:42Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"sP4PSsNmizXXeq9f2HRr4fbNbSJIUGU4HOAv7JW6xMDUTn39z8//kTkyMi0fww6qDOsIZH14acQC+aAUN9DvCw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-07-07T11:17:15.808633Z"},"content_sha256":"61521605070b94e258ec84581d04897497d21bf79c83c74a6ba4932baea73dd2","schema_version":"1.0","event_id":"sha256:61521605070b94e258ec84581d04897497d21bf79c83c74a6ba4932baea73dd2"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/7KFCMEZLG2AU6FGYFTHZQMEQ6E/bundle.json","state_url":"https://pith.science/pith/7KFCMEZLG2AU6FGYFTHZQMEQ6E/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/7KFCMEZLG2AU6FGYFTHZQMEQ6E/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-07-07T11:17:15Z","links":{"resolver":"https://pith.science/pith/7KFCMEZLG2AU6FGYFTHZQMEQ6E","bundle":"https://pith.science/pith/7KFCMEZLG2AU6FGYFTHZQMEQ6E/bundle.json","state":"https://pith.science/pith/7KFCMEZLG2AU6FGYFTHZQMEQ6E/state.json","well_known_bundle":"https://pith.science/.well-known/pith/7KFCMEZLG2AU6FGYFTHZQMEQ6E/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2024:7KFCMEZLG2AU6FGYFTHZQMEQ6E","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"3e60ea14c28940bd3776853c95e01b6bdfcafb8750549fd3033a02895eb20075","cross_cats_sorted":["cs.AI","cs.LG"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-07-16T17:59:55Z","title_canon_sha256":"2e07a7126588b9b4682dc080eeaed96734570bc22985626ede93f0185a804058"},"schema_version":"1.0","source":{"id":"2407.11969","kind":"arxiv","version":4}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2407.11969","created_at":"2026-07-05T10:50:42Z"},{"alias_kind":"arxiv_version","alias_value":"2407.11969v4","created_at":"2026-07-05T10:50:42Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2407.11969","created_at":"2026-07-05T10:50:42Z"},{"alias_kind":"pith_short_12","alias_value":"7KFCMEZLG2AU","created_at":"2026-07-05T10:50:42Z"},{"alias_kind":"pith_short_16","alias_value":"7KFCMEZLG2AU6FGY","created_at":"2026-07-05T10:50:42Z"},{"alias_kind":"pith_short_8","alias_value":"7KFCMEZL","created_at":"2026-07-05T10:50:42Z"}],"graph_snapshots":[{"event_id":"sha256:61521605070b94e258ec84581d04897497d21bf79c83c74a6ba4932baea73dd2","target":"graph","created_at":"2026-07-05T10:50:42Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2407.11969/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Refusal training is widely used to prevent LLMs from generating harmful, undesirable, or illegal outputs. We reveal a curious generalization gap in the current refusal training approaches: simply reformulating a harmful request in the past tense (e.g., \"How to make a Molotov cocktail?\" to \"How did people make a Molotov cocktail?\") is often sufficient to jailbreak many state-of-the-art LLMs. We systematically evaluate this method on Llama-3 8B, Claude-3.5 Sonnet, GPT-3.5 Turbo, Gemma-2 9B, Phi-3-Mini, GPT-4o mini, GPT-4o, o1-mini, o1-preview, and R2D2 models using GPT-3.5 Turbo as a reformulati","authors_text":"Maksym Andriushchenko, Nicolas Flammarion","cross_cats":["cs.AI","cs.LG"],"headline":"","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-07-16T17:59:55Z","title":"Does Refusal Training in LLMs Generalize to the Past Tense?"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2407.11969","kind":"arxiv","version":4},"verdict":{"created_at":null,"id":null,"model_set":{},"one_line_summary":"","pipeline_version":null,"pith_extraction_headline":"","strongest_claim":"","weakest_assumption":""}},"verdict_id":null}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:838e68b9a18fcdd632cd4856245f480b605d4d6a85608dd37b845d714b744f5e","target":"record","created_at":"2026-07-05T10:50:42Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"3e60ea14c28940bd3776853c95e01b6bdfcafb8750549fd3033a02895eb20075","cross_cats_sorted":["cs.AI","cs.LG"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-07-16T17:59:55Z","title_canon_sha256":"2e07a7126588b9b4682dc080eeaed96734570bc22985626ede93f0185a804058"},"schema_version":"1.0","source":{"id":"2407.11969","kind":"arxiv","version":4}},"canonical_sha256":"fa8a26132b36814f14d82ccf983090f1387bcb8ffd5e4a713c979345aed5160a","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"fa8a26132b36814f14d82ccf983090f1387bcb8ffd5e4a713c979345aed5160a","first_computed_at":"2026-07-05T10:50:42.938798Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T10:50:42.938798Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"eoQ/4f7tBxbXjuTmHg8GYsaHyhlrT4hCodIUADWk4Yj8usNdw+l9ArqKC0MITQAlfVEZXF9CT91aHxNkBfmYCA==","signature_status":"signed_v1","signed_at":"2026-07-05T10:50:42.939304Z","signed_message":"canonical_sha256_bytes"},"source_id":"2407.11969","source_kind":"arxiv","source_version":4}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:838e68b9a18fcdd632cd4856245f480b605d4d6a85608dd37b845d714b744f5e","sha256:61521605070b94e258ec84581d04897497d21bf79c83c74a6ba4932baea73dd2"],"state_sha256":"c1371fe28262060f62dd838e59040008798d68c89384a72668804e2eb1878646"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"KPMVKXeYOKsYI6ArdI1LtoJo0b7qNX8p+gjJs1kMpwhVS9GMryagX4KWZOMrQyI53sNHcBLIIi0AazOPO7byDA==","signed_message":"bundle_sha256_bytes","signed_at":"2026-07-07T11:17:15.810623Z","bundle_sha256":"100b15a22a26a9ebc765983c4755b8be3a89392409510c012ebae583ff185bfa"}}