{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2024:KLF577JGEJPVT7UFVL6ELCNMSB","short_pith_number":"pith:KLF577JG","canonical_record":{"source":{"id":"2404.02151","kind":"arxiv","version":4},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CR","submitted_at":"2024-04-02T17:58:27Z","cross_cats_sorted":["cs.AI","cs.LG","stat.ML"],"title_canon_sha256":"b2110790cec662d758eacac98b117aa1d1e3adb5fcee666601d53183e271eb17","abstract_canon_sha256":"707d49577db88f0eb53e78087ba8a601bb6da184c14c96ed39fef320909149ce"},"schema_version":"1.0"},"canonical_sha256":"52cbdffd26225f59fe85aafc4589ac90656b3a7ac643e2b5d14a169c8d20fa71","source":{"kind":"arxiv","id":"2404.02151","version":4},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2404.02151","created_at":"2026-07-05T10:50:41Z"},{"alias_kind":"arxiv_version","alias_value":"2404.02151v4","created_at":"2026-07-05T10:50:41Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2404.02151","created_at":"2026-07-05T10:50:41Z"},{"alias_kind":"pith_short_12","alias_value":"KLF577JGEJPV","created_at":"2026-07-05T10:50:41Z"},{"alias_kind":"pith_short_16","alias_value":"KLF577JGEJPVT7UF","created_at":"2026-07-05T10:50:41Z"},{"alias_kind":"pith_short_8","alias_value":"KLF577JG","created_at":"2026-07-05T10:50:41Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2024:KLF577JGEJPVT7UFVL6ELCNMSB","target":"record","payload":{"canonical_record":{"source":{"id":"2404.02151","kind":"arxiv","version":4},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CR","submitted_at":"2024-04-02T17:58:27Z","cross_cats_sorted":["cs.AI","cs.LG","stat.ML"],"title_canon_sha256":"b2110790cec662d758eacac98b117aa1d1e3adb5fcee666601d53183e271eb17","abstract_canon_sha256":"707d49577db88f0eb53e78087ba8a601bb6da184c14c96ed39fef320909149ce"},"schema_version":"1.0"},"canonical_sha256":"52cbdffd26225f59fe85aafc4589ac90656b3a7ac643e2b5d14a169c8d20fa71","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:50:41.945170Z","signature_b64":"WQo/n0ZbmF4qM62e5eADh2sw8jlo852m7cjaViXyxrTKyPW4YHXd6Y1WdOTsf868arZXjU+ZV8oLOS4z/KDOBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"52cbdffd26225f59fe85aafc4589ac90656b3a7ac643e2b5d14a169c8d20fa71","last_reissued_at":"2026-07-05T10:50:41.944737Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:50:41.944737Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2404.02151","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":"2026-07-05T10:50:41Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"tKlx74vbD6jfqvAkTMw1ReWTGWAc+Njj4StwX6l9AojzWW0p9d/PlNGmAaE9WyfEud9D/Ri6CXkY0HOQE6+KDg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-07-06T18:47:27.084611Z"},"content_sha256":"30c664c59457d6748055f156d934162a15dd8467b9f55ccb9ba5cf73945771fe","schema_version":"1.0","event_id":"sha256:30c664c59457d6748055f156d934162a15dd8467b9f55ccb9ba5cf73945771fe"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2024:KLF577JGEJPVT7UFVL6ELCNMSB","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Jailbreaking Leading Safety-Aligned LLMs with Simple Adaptive Attacks","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG","stat.ML"],"primary_cat":"cs.CR","authors_text":"Francesco Croce, Maksym Andriushchenko, Nicolas Flammarion","submitted_at":"2024-04-02T17:58:27Z","abstract_excerpt":"We show that even the most recent safety-aligned LLMs are not robust to simple adaptive jailbreaking attacks. First, we demonstrate how to successfully leverage access to logprobs for jailbreaking: we initially design an adversarial prompt template (sometimes adapted to the target LLM), and then we apply random search on a suffix to maximize a target logprob (e.g., of the token \"Sure\"), potentially with multiple restarts. In this way, we achieve 100% attack success rate -- according to GPT-4 as a judge -- on Vicuna-13B, Mistral-7B, Phi-3-Mini, Nemotron-4-340B, Llama-2-Chat-7B/13B/70B, Llama-3-"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2404.02151","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/2404.02151/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":"2026-07-05T10:50:41Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"S6hvVj35YqG1uRAAWGuDviV+jZqFjX546gQh0AMuwdAYy7wVDtGLH2DrMhFnId7L9E8FLL4mFlePfFKatcnGCQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-07-06T18:47:27.084982Z"},"content_sha256":"bacceddf6f9285b217a6bab191f4cdd7b85aa56bbf25aee14605255692ea9769","schema_version":"1.0","event_id":"sha256:bacceddf6f9285b217a6bab191f4cdd7b85aa56bbf25aee14605255692ea9769"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/KLF577JGEJPVT7UFVL6ELCNMSB/bundle.json","state_url":"https://pith.science/pith/KLF577JGEJPVT7UFVL6ELCNMSB/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/KLF577JGEJPVT7UFVL6ELCNMSB/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":"2026-07-06T18:47:27Z","links":{"resolver":"https://pith.science/pith/KLF577JGEJPVT7UFVL6ELCNMSB","bundle":"https://pith.science/pith/KLF577JGEJPVT7UFVL6ELCNMSB/bundle.json","state":"https://pith.science/pith/KLF577JGEJPVT7UFVL6ELCNMSB/state.json","well_known_bundle":"https://pith.science/.well-known/pith/KLF577JGEJPVT7UFVL6ELCNMSB/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2024:KLF577JGEJPVT7UFVL6ELCNMSB","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":"707d49577db88f0eb53e78087ba8a601bb6da184c14c96ed39fef320909149ce","cross_cats_sorted":["cs.AI","cs.LG","stat.ML"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CR","submitted_at":"2024-04-02T17:58:27Z","title_canon_sha256":"b2110790cec662d758eacac98b117aa1d1e3adb5fcee666601d53183e271eb17"},"schema_version":"1.0","source":{"id":"2404.02151","kind":"arxiv","version":4}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2404.02151","created_at":"2026-07-05T10:50:41Z"},{"alias_kind":"arxiv_version","alias_value":"2404.02151v4","created_at":"2026-07-05T10:50:41Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2404.02151","created_at":"2026-07-05T10:50:41Z"},{"alias_kind":"pith_short_12","alias_value":"KLF577JGEJPV","created_at":"2026-07-05T10:50:41Z"},{"alias_kind":"pith_short_16","alias_value":"KLF577JGEJPVT7UF","created_at":"2026-07-05T10:50:41Z"},{"alias_kind":"pith_short_8","alias_value":"KLF577JG","created_at":"2026-07-05T10:50:41Z"}],"graph_snapshots":[{"event_id":"sha256:bacceddf6f9285b217a6bab191f4cdd7b85aa56bbf25aee14605255692ea9769","target":"graph","created_at":"2026-07-05T10:50:41Z","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/2404.02151/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"We show that even the most recent safety-aligned LLMs are not robust to simple adaptive jailbreaking attacks. First, we demonstrate how to successfully leverage access to logprobs for jailbreaking: we initially design an adversarial prompt template (sometimes adapted to the target LLM), and then we apply random search on a suffix to maximize a target logprob (e.g., of the token \"Sure\"), potentially with multiple restarts. In this way, we achieve 100% attack success rate -- according to GPT-4 as a judge -- on Vicuna-13B, Mistral-7B, Phi-3-Mini, Nemotron-4-340B, Llama-2-Chat-7B/13B/70B, Llama-3-","authors_text":"Francesco Croce, Maksym Andriushchenko, Nicolas Flammarion","cross_cats":["cs.AI","cs.LG","stat.ML"],"headline":"","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CR","submitted_at":"2024-04-02T17:58:27Z","title":"Jailbreaking Leading Safety-Aligned LLMs with Simple Adaptive Attacks"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2404.02151","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":"sha256:30c664c59457d6748055f156d934162a15dd8467b9f55ccb9ba5cf73945771fe","target":"record","created_at":"2026-07-05T10:50:41Z","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":"707d49577db88f0eb53e78087ba8a601bb6da184c14c96ed39fef320909149ce","cross_cats_sorted":["cs.AI","cs.LG","stat.ML"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CR","submitted_at":"2024-04-02T17:58:27Z","title_canon_sha256":"b2110790cec662d758eacac98b117aa1d1e3adb5fcee666601d53183e271eb17"},"schema_version":"1.0","source":{"id":"2404.02151","kind":"arxiv","version":4}},"canonical_sha256":"52cbdffd26225f59fe85aafc4589ac90656b3a7ac643e2b5d14a169c8d20fa71","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"52cbdffd26225f59fe85aafc4589ac90656b3a7ac643e2b5d14a169c8d20fa71","first_computed_at":"2026-07-05T10:50:41.944737Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T10:50:41.944737Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"WQo/n0ZbmF4qM62e5eADh2sw8jlo852m7cjaViXyxrTKyPW4YHXd6Y1WdOTsf868arZXjU+ZV8oLOS4z/KDOBw==","signature_status":"signed_v1","signed_at":"2026-07-05T10:50:41.945170Z","signed_message":"canonical_sha256_bytes"},"source_id":"2404.02151","source_kind":"arxiv","source_version":4}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:30c664c59457d6748055f156d934162a15dd8467b9f55ccb9ba5cf73945771fe","sha256:bacceddf6f9285b217a6bab191f4cdd7b85aa56bbf25aee14605255692ea9769"],"state_sha256":"118af0075c336d01ac8d90ff8184a9727811b193bf6f3a439d8247e35a03bd32"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"Igr2vqfRcxQ0mNPyN/o5wteEKbYYHuSSjEN1BY7NDVR9iDvlpO8EyGjUs8/paE06+crzOjrSjFCLPzrwv6zcAg==","signed_message":"bundle_sha256_bytes","signed_at":"2026-07-06T18:47:27.087095Z","bundle_sha256":"ffcccb56b579e6f718b932ac135218845e4dca82883c0c7fe93d6959625b2bb7"}}