{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2023:UTUUUBAYBZE6H3L7OVXI3FVVUU","short_pith_number":"pith:UTUUUBAY","canonical_record":{"source":{"id":"2309.07875","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2023-09-14T17:23:37Z","cross_cats_sorted":[],"title_canon_sha256":"225f72aa6efd61771c5d0f70a82064571b12f31a83993a1ec2afc94ba18ed885","abstract_canon_sha256":"66602fe3b9cb8a0cd37bafcb30f4bf383a2ba583d2d8b267df1075a7ed4e9162"},"schema_version":"1.0"},"canonical_sha256":"a4e94a04180e49e3ed7f756e8d96b5a519e114b82ea0e26b98a3cd24e827079b","source":{"kind":"arxiv","id":"2309.07875","version":3},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2309.07875","created_at":"2026-07-05T07:57:43Z"},{"alias_kind":"arxiv_version","alias_value":"2309.07875v3","created_at":"2026-07-05T07:57:43Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2309.07875","created_at":"2026-07-05T07:57:43Z"},{"alias_kind":"pith_short_12","alias_value":"UTUUUBAYBZE6","created_at":"2026-07-05T07:57:43Z"},{"alias_kind":"pith_short_16","alias_value":"UTUUUBAYBZE6H3L7","created_at":"2026-07-05T07:57:43Z"},{"alias_kind":"pith_short_8","alias_value":"UTUUUBAY","created_at":"2026-07-05T07:57:43Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2023:UTUUUBAYBZE6H3L7OVXI3FVVUU","target":"record","payload":{"canonical_record":{"source":{"id":"2309.07875","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2023-09-14T17:23:37Z","cross_cats_sorted":[],"title_canon_sha256":"225f72aa6efd61771c5d0f70a82064571b12f31a83993a1ec2afc94ba18ed885","abstract_canon_sha256":"66602fe3b9cb8a0cd37bafcb30f4bf383a2ba583d2d8b267df1075a7ed4e9162"},"schema_version":"1.0"},"canonical_sha256":"a4e94a04180e49e3ed7f756e8d96b5a519e114b82ea0e26b98a3cd24e827079b","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:57:43.614463Z","signature_b64":"MKe2G5Uuh6A2M3oDsGpU5tMKaIXUB0dgoeH2TQned/V34fc6Z5B5TN/09uqqifJ+HTHdWj/3nxD5NQLKsllWAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a4e94a04180e49e3ed7f756e8d96b5a519e114b82ea0e26b98a3cd24e827079b","last_reissued_at":"2026-07-05T07:57:43.613961Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:57:43.613961Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2309.07875","source_version":3,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T07:57:43Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"EL5feXs0zgSW3ozoKr/ol3Bk3uTVC37yJu7OK6dTHc8ADuF2aJEKBZhcEUUBL1SGiuU1W+oB0KoH14cuLehrAQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-03T19:20:56.168374Z"},"content_sha256":"84d84298a718efb994dc7442639ee5ff533c4f7e7aea041069dc7d7d57c88ff0","schema_version":"1.0","event_id":"sha256:84d84298a718efb994dc7442639ee5ff533c4f7e7aea041069dc7d7d57c88ff0"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2023:UTUUUBAYBZE6H3L7OVXI3FVVUU","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Safety-Tuned LLaMAs: Lessons From Improving the Safety of Large Language Models that Follow Instructions","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Dan Jurafsky, Federico Bianchi, Giuseppe Attanasio, James Zou, Mirac Suzgun, Paul R\\\"ottger, Tatsunori Hashimoto","submitted_at":"2023-09-14T17:23:37Z","abstract_excerpt":"Training large language models to follow instructions makes them perform better on a wide range of tasks and generally become more helpful. However, a perfectly helpful model will follow even the most malicious instructions and readily generate harmful content. In this paper, we raise concerns over the safety of models that only emphasize helpfulness, not harmlessness, in their instruction-tuning. We show that several popular instruction-tuned models are highly unsafe. Moreover, we show that adding just 3% safety examples (a few hundred demonstrations) when fine-tuning a model like LLaMA can s"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2309.07875","kind":"arxiv","version":3},"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/2309.07875/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":"2026-07-05T07:57:43Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"j7E2hg4XJjSIfW0SmUF+8T0S1F8Mb9nNSxTSV/IxEWiYVEHH6/LLYAEfy4NPR9E047j2vMD4x5iX0gDGEs4/Cg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-03T19:20:56.169390Z"},"content_sha256":"4c7700cc227fcadb525b30fc21fa689d90a5b62c7953bd897842bd2aba54fd9e","schema_version":"1.0","event_id":"sha256:4c7700cc227fcadb525b30fc21fa689d90a5b62c7953bd897842bd2aba54fd9e"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/UTUUUBAYBZE6H3L7OVXI3FVVUU/bundle.json","state_url":"https://pith.science/pith/UTUUUBAYBZE6H3L7OVXI3FVVUU/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/UTUUUBAYBZE6H3L7OVXI3FVVUU/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":"2026-08-03T19:20:56Z","links":{"resolver":"https://pith.science/pith/UTUUUBAYBZE6H3L7OVXI3FVVUU","bundle":"https://pith.science/pith/UTUUUBAYBZE6H3L7OVXI3FVVUU/bundle.json","state":"https://pith.science/pith/UTUUUBAYBZE6H3L7OVXI3FVVUU/state.json","well_known_bundle":"https://pith.science/.well-known/pith/UTUUUBAYBZE6H3L7OVXI3FVVUU/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2023:UTUUUBAYBZE6H3L7OVXI3FVVUU","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":"66602fe3b9cb8a0cd37bafcb30f4bf383a2ba583d2d8b267df1075a7ed4e9162","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2023-09-14T17:23:37Z","title_canon_sha256":"225f72aa6efd61771c5d0f70a82064571b12f31a83993a1ec2afc94ba18ed885"},"schema_version":"1.0","source":{"id":"2309.07875","kind":"arxiv","version":3}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2309.07875","created_at":"2026-07-05T07:57:43Z"},{"alias_kind":"arxiv_version","alias_value":"2309.07875v3","created_at":"2026-07-05T07:57:43Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2309.07875","created_at":"2026-07-05T07:57:43Z"},{"alias_kind":"pith_short_12","alias_value":"UTUUUBAYBZE6","created_at":"2026-07-05T07:57:43Z"},{"alias_kind":"pith_short_16","alias_value":"UTUUUBAYBZE6H3L7","created_at":"2026-07-05T07:57:43Z"},{"alias_kind":"pith_short_8","alias_value":"UTUUUBAY","created_at":"2026-07-05T07:57:43Z"}],"graph_snapshots":[{"event_id":"sha256:4c7700cc227fcadb525b30fc21fa689d90a5b62c7953bd897842bd2aba54fd9e","target":"graph","created_at":"2026-07-05T07:57:43Z","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/2309.07875/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Training large language models to follow instructions makes them perform better on a wide range of tasks and generally become more helpful. However, a perfectly helpful model will follow even the most malicious instructions and readily generate harmful content. In this paper, we raise concerns over the safety of models that only emphasize helpfulness, not harmlessness, in their instruction-tuning. We show that several popular instruction-tuned models are highly unsafe. Moreover, we show that adding just 3% safety examples (a few hundred demonstrations) when fine-tuning a model like LLaMA can s","authors_text":"Dan Jurafsky, Federico Bianchi, Giuseppe Attanasio, James Zou, Mirac Suzgun, Paul R\\\"ottger, Tatsunori Hashimoto","cross_cats":[],"headline":"","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2023-09-14T17:23:37Z","title":"Safety-Tuned LLaMAs: Lessons From Improving the Safety of Large Language Models that Follow Instructions"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2309.07875","kind":"arxiv","version":3},"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":"sha256:84d84298a718efb994dc7442639ee5ff533c4f7e7aea041069dc7d7d57c88ff0","target":"record","created_at":"2026-07-05T07:57:43Z","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":"66602fe3b9cb8a0cd37bafcb30f4bf383a2ba583d2d8b267df1075a7ed4e9162","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2023-09-14T17:23:37Z","title_canon_sha256":"225f72aa6efd61771c5d0f70a82064571b12f31a83993a1ec2afc94ba18ed885"},"schema_version":"1.0","source":{"id":"2309.07875","kind":"arxiv","version":3}},"canonical_sha256":"a4e94a04180e49e3ed7f756e8d96b5a519e114b82ea0e26b98a3cd24e827079b","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"a4e94a04180e49e3ed7f756e8d96b5a519e114b82ea0e26b98a3cd24e827079b","first_computed_at":"2026-07-05T07:57:43.613961Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T07:57:43.613961Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"MKe2G5Uuh6A2M3oDsGpU5tMKaIXUB0dgoeH2TQned/V34fc6Z5B5TN/09uqqifJ+HTHdWj/3nxD5NQLKsllWAQ==","signature_status":"signed_v1","signed_at":"2026-07-05T07:57:43.614463Z","signed_message":"canonical_sha256_bytes"},"source_id":"2309.07875","source_kind":"arxiv","source_version":3}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:84d84298a718efb994dc7442639ee5ff533c4f7e7aea041069dc7d7d57c88ff0","sha256:4c7700cc227fcadb525b30fc21fa689d90a5b62c7953bd897842bd2aba54fd9e"],"state_sha256":"7fa6e8c2a94bdac7e1fe09f70ec90a28cf05a676c33867fc8fbcb71b4e993f89"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"llIHPaHmm1P8+eXcbqRN2ZwTnQ+tTZM03vVVaSO1nmA3vF11d0t1PgBZ4ftWi3OL2d5gv5vpD3FSroZcDlL4Bw==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-03T19:20:56.178725Z","bundle_sha256":"1b265bfa5de4293e27a187176e61c8c314a0373f73fdf65944197f3eecad4a74"}}