{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2023:XWSBVZBK5G524AZ3T63S6EYV4U","short_pith_number":"pith:XWSBVZBK","canonical_record":{"source":{"id":"2311.07689","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2023-11-13T19:13:29Z","cross_cats_sorted":[],"title_canon_sha256":"7ebc4247e6fd0940dbce7b0b3258bf3f805eb7e88b4ea1a2995a8757a0d75eb9","abstract_canon_sha256":"51584e8daa17c65adf7686551895ba921b7dcdb973523ed15de25167ce971953"},"schema_version":"1.0"},"canonical_sha256":"bda41ae42ae9bbae033b9fb72f1315e518ae8f3e5dd452a8570d546dc76d60d5","source":{"kind":"arxiv","id":"2311.07689","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2311.07689","created_at":"2026-07-05T07:12:22Z"},{"alias_kind":"arxiv_version","alias_value":"2311.07689v1","created_at":"2026-07-05T07:12:22Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2311.07689","created_at":"2026-07-05T07:12:22Z"},{"alias_kind":"pith_short_12","alias_value":"XWSBVZBK5G52","created_at":"2026-07-05T07:12:22Z"},{"alias_kind":"pith_short_16","alias_value":"XWSBVZBK5G524AZ3","created_at":"2026-07-05T07:12:22Z"},{"alias_kind":"pith_short_8","alias_value":"XWSBVZBK","created_at":"2026-07-05T07:12:22Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2023:XWSBVZBK5G524AZ3T63S6EYV4U","target":"record","payload":{"canonical_record":{"source":{"id":"2311.07689","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2023-11-13T19:13:29Z","cross_cats_sorted":[],"title_canon_sha256":"7ebc4247e6fd0940dbce7b0b3258bf3f805eb7e88b4ea1a2995a8757a0d75eb9","abstract_canon_sha256":"51584e8daa17c65adf7686551895ba921b7dcdb973523ed15de25167ce971953"},"schema_version":"1.0"},"canonical_sha256":"bda41ae42ae9bbae033b9fb72f1315e518ae8f3e5dd452a8570d546dc76d60d5","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:12:22.420891Z","signature_b64":"VS/NVFh2DADeShIgGq9LJ85iEcocbY/UzK+KHAQqrg1buWRuPceIQ1anWeUcJCjpKlYm/knDWyG4G23GMKn6CQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"bda41ae42ae9bbae033b9fb72f1315e518ae8f3e5dd452a8570d546dc76d60d5","last_reissued_at":"2026-07-05T07:12:22.420448Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:12:22.420448Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2311.07689","source_version":1,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T07:12:22Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"o/ujlmXZnQHNR8zfW8A+hgYL9TPb3C32tCZ9ySNWsogFg2xIm52vSpzGaM2MqomyZ+9cC81w690ajn+in9IoBg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-10T08:43:15.916077Z"},"content_sha256":"56295a55e72d38029430c81662d8fb939a62c7b203a1267a073d66bd6821891f","schema_version":"1.0","event_id":"sha256:56295a55e72d38029430c81662d8fb939a62c7b203a1267a073d66bd6821891f"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2023:XWSBVZBK5G524AZ3T63S6EYV4U","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"MART: Improving LLM Safety with Multi-round Automatic Red-Teaming","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Chunting Zhou, Jiawei Han, Madian Khabsa, Qifan Wang, Rui Hou, Suyu Ge, Yi-Chia Wang, Yuning Mao","submitted_at":"2023-11-13T19:13:29Z","abstract_excerpt":"Red-teaming is a common practice for mitigating unsafe behaviors in Large Language Models (LLMs), which involves thoroughly assessing LLMs to identify potential flaws and addressing them with responsible and accurate responses. While effective, manual red-teaming is costly, and existing automatic red-teaming typically discovers safety risks without addressing them. In this paper, we propose a Multi-round Automatic Red-Teaming (MART) method, which incorporates both automatic adversarial prompt writing and safe response generation, significantly increasing red-teaming scalability and the safety "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2311.07689","kind":"arxiv","version":1},"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/2311.07689/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":"2026-07-05T07:12:22Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"fJND6QHAGtPkJ09ZTL9eTm4IW3u5rAZkLhxbLHJPkTO7COmqeu1a9e+54t/UbpUtXa2YHm0zLmiWRYuzh3JJDg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-10T08:43:15.916966Z"},"content_sha256":"4530a6c6e206dcf458d40873d6dbde7e8d7cec53f19a89262a2eeaf39bf2f0a9","schema_version":"1.0","event_id":"sha256:4530a6c6e206dcf458d40873d6dbde7e8d7cec53f19a89262a2eeaf39bf2f0a9"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/XWSBVZBK5G524AZ3T63S6EYV4U/bundle.json","state_url":"https://pith.science/pith/XWSBVZBK5G524AZ3T63S6EYV4U/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/XWSBVZBK5G524AZ3T63S6EYV4U/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":"2026-08-10T08:43:15Z","links":{"resolver":"https://pith.science/pith/XWSBVZBK5G524AZ3T63S6EYV4U","bundle":"https://pith.science/pith/XWSBVZBK5G524AZ3T63S6EYV4U/bundle.json","state":"https://pith.science/pith/XWSBVZBK5G524AZ3T63S6EYV4U/state.json","well_known_bundle":"https://pith.science/.well-known/pith/XWSBVZBK5G524AZ3T63S6EYV4U/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2023:XWSBVZBK5G524AZ3T63S6EYV4U","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":"51584e8daa17c65adf7686551895ba921b7dcdb973523ed15de25167ce971953","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2023-11-13T19:13:29Z","title_canon_sha256":"7ebc4247e6fd0940dbce7b0b3258bf3f805eb7e88b4ea1a2995a8757a0d75eb9"},"schema_version":"1.0","source":{"id":"2311.07689","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2311.07689","created_at":"2026-07-05T07:12:22Z"},{"alias_kind":"arxiv_version","alias_value":"2311.07689v1","created_at":"2026-07-05T07:12:22Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2311.07689","created_at":"2026-07-05T07:12:22Z"},{"alias_kind":"pith_short_12","alias_value":"XWSBVZBK5G52","created_at":"2026-07-05T07:12:22Z"},{"alias_kind":"pith_short_16","alias_value":"XWSBVZBK5G524AZ3","created_at":"2026-07-05T07:12:22Z"},{"alias_kind":"pith_short_8","alias_value":"XWSBVZBK","created_at":"2026-07-05T07:12:22Z"}],"graph_snapshots":[{"event_id":"sha256:4530a6c6e206dcf458d40873d6dbde7e8d7cec53f19a89262a2eeaf39bf2f0a9","target":"graph","created_at":"2026-07-05T07:12:22Z","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/2311.07689/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Red-teaming is a common practice for mitigating unsafe behaviors in Large Language Models (LLMs), which involves thoroughly assessing LLMs to identify potential flaws and addressing them with responsible and accurate responses. While effective, manual red-teaming is costly, and existing automatic red-teaming typically discovers safety risks without addressing them. In this paper, we propose a Multi-round Automatic Red-Teaming (MART) method, which incorporates both automatic adversarial prompt writing and safe response generation, significantly increasing red-teaming scalability and the safety ","authors_text":"Chunting Zhou, Jiawei Han, Madian Khabsa, Qifan Wang, Rui Hou, Suyu Ge, Yi-Chia Wang, Yuning Mao","cross_cats":[],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2023-11-13T19:13:29Z","title":"MART: Improving LLM Safety with Multi-round Automatic Red-Teaming"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2311.07689","kind":"arxiv","version":1},"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":"sha256:56295a55e72d38029430c81662d8fb939a62c7b203a1267a073d66bd6821891f","target":"record","created_at":"2026-07-05T07:12:22Z","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":"51584e8daa17c65adf7686551895ba921b7dcdb973523ed15de25167ce971953","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2023-11-13T19:13:29Z","title_canon_sha256":"7ebc4247e6fd0940dbce7b0b3258bf3f805eb7e88b4ea1a2995a8757a0d75eb9"},"schema_version":"1.0","source":{"id":"2311.07689","kind":"arxiv","version":1}},"canonical_sha256":"bda41ae42ae9bbae033b9fb72f1315e518ae8f3e5dd452a8570d546dc76d60d5","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"bda41ae42ae9bbae033b9fb72f1315e518ae8f3e5dd452a8570d546dc76d60d5","first_computed_at":"2026-07-05T07:12:22.420448Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T07:12:22.420448Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"VS/NVFh2DADeShIgGq9LJ85iEcocbY/UzK+KHAQqrg1buWRuPceIQ1anWeUcJCjpKlYm/knDWyG4G23GMKn6CQ==","signature_status":"signed_v1","signed_at":"2026-07-05T07:12:22.420891Z","signed_message":"canonical_sha256_bytes"},"source_id":"2311.07689","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:56295a55e72d38029430c81662d8fb939a62c7b203a1267a073d66bd6821891f","sha256:4530a6c6e206dcf458d40873d6dbde7e8d7cec53f19a89262a2eeaf39bf2f0a9"],"state_sha256":"e4e42face9e4443867804cff876b29237b2e418252c843e5d7477977a85a9aaf"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"bQ6liRClRZNLUv/iOecshxocyZVzw2bPH4tFWXg6ANGaxE4jo9qAS8VRvoXUfeLX0xSsrhQMya5P7Iu+QtI2Bg==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-10T08:43:15.923935Z","bundle_sha256":"51255edf7019678ce35aede2e7eee1146ced112647fd5ab5d0af7bef18d736c9"}}