{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2024:GGXWKXV6A6KQ5MMELZ22KYFXMP","short_pith_number":"pith:GGXWKXV6","canonical_record":{"source":{"id":"2402.18540","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-02-28T18:23:49Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"92c76a4ca142260c092afea071ddc39d118d5471cf5ea303165af94fa5f97dd0","abstract_canon_sha256":"711a9bd39d6e66517ee976a6d6309caba66c335fd5d6375860072de44e1e9740"},"schema_version":"1.0"},"canonical_sha256":"31af655ebe07950eb1845e75a560b763df8d7cda9b8c2ae03726f9e843d0cb16","source":{"kind":"arxiv","id":"2402.18540","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2402.18540","created_at":"2026-07-05T10:02:02Z"},{"alias_kind":"arxiv_version","alias_value":"2402.18540v2","created_at":"2026-07-05T10:02:02Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2402.18540","created_at":"2026-07-05T10:02:02Z"},{"alias_kind":"pith_short_12","alias_value":"GGXWKXV6A6KQ","created_at":"2026-07-05T10:02:02Z"},{"alias_kind":"pith_short_16","alias_value":"GGXWKXV6A6KQ5MME","created_at":"2026-07-05T10:02:02Z"},{"alias_kind":"pith_short_8","alias_value":"GGXWKXV6","created_at":"2026-07-05T10:02:02Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2024:GGXWKXV6A6KQ5MMELZ22KYFXMP","target":"record","payload":{"canonical_record":{"source":{"id":"2402.18540","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-02-28T18:23:49Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"92c76a4ca142260c092afea071ddc39d118d5471cf5ea303165af94fa5f97dd0","abstract_canon_sha256":"711a9bd39d6e66517ee976a6d6309caba66c335fd5d6375860072de44e1e9740"},"schema_version":"1.0"},"canonical_sha256":"31af655ebe07950eb1845e75a560b763df8d7cda9b8c2ae03726f9e843d0cb16","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:02:02.303013Z","signature_b64":"qCzUXp8bnZNi3n6jhxeU1yG6w1pWNqNpWntsBZos2qy/l+3rDdjdouh4m9Y0oAezuf7WlLCf14G9hd5tmqWGDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"31af655ebe07950eb1845e75a560b763df8d7cda9b8c2ae03726f9e843d0cb16","last_reissued_at":"2026-07-05T10:02:02.302594Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:02:02.302594Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2402.18540","source_version":2,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T10:02:02Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"kkbIcKJa71Z12+YaOcelIXpSagIuuIaG9Y1bRJ8KF5FgNRyKZPeu1p0WAiSqCwiPjJd0LdaIFrKIAOfDsgEHDg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-08T13:56:09.145609Z"},"content_sha256":"f2c8ff06f7d2cd733fc800342212e504b01131cc417e52177c38a066b0d71458","schema_version":"1.0","event_id":"sha256:f2c8ff06f7d2cd733fc800342212e504b01131cc417e52177c38a066b0d71458"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2024:GGXWKXV6A6KQ5MMELZ22KYFXMP","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Keeping LLMs Aligned After Fine-tuning: The Crucial Role of Prompt Templates","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Anirudh Goyal, Dingli Yu, Haoyu Zhao, Kaifeng Lyu, Sanjeev Arora, Xinran Gu","submitted_at":"2024-02-28T18:23:49Z","abstract_excerpt":"Public LLMs such as the Llama 2-Chat underwent alignment training and were considered safe. Recently Qi et al. [2024] reported that even benign fine-tuning on seemingly safe datasets can give rise to unsafe behaviors in the models. The current paper is about methods and best practices to mitigate such loss of alignment. We focus on the setting where a public model is fine-tuned before serving users for specific usage, where the model should improve on the downstream task while maintaining alignment. Through extensive experiments on several chat models (Meta's Llama 2-Chat, Mistral AI's Mistral"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2402.18540","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/2402.18540/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":"2026-07-05T10:02:02Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"Mp285SRcgBtCVKoycZo0TzPtlv+XAUBBGB9uhVPy7g/xIuIfQYqIKh4qwKuuKbE1CKOH8cfYmabJyNeUWx8eBQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-08T13:56:09.146120Z"},"content_sha256":"c0b20d371c6afd230abcbbc2c555e2988b12c852f6b10c19a2665d83dc0d87da","schema_version":"1.0","event_id":"sha256:c0b20d371c6afd230abcbbc2c555e2988b12c852f6b10c19a2665d83dc0d87da"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/GGXWKXV6A6KQ5MMELZ22KYFXMP/bundle.json","state_url":"https://pith.science/pith/GGXWKXV6A6KQ5MMELZ22KYFXMP/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/GGXWKXV6A6KQ5MMELZ22KYFXMP/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":"2026-08-08T13:56:09Z","links":{"resolver":"https://pith.science/pith/GGXWKXV6A6KQ5MMELZ22KYFXMP","bundle":"https://pith.science/pith/GGXWKXV6A6KQ5MMELZ22KYFXMP/bundle.json","state":"https://pith.science/pith/GGXWKXV6A6KQ5MMELZ22KYFXMP/state.json","well_known_bundle":"https://pith.science/.well-known/pith/GGXWKXV6A6KQ5MMELZ22KYFXMP/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2024:GGXWKXV6A6KQ5MMELZ22KYFXMP","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":"711a9bd39d6e66517ee976a6d6309caba66c335fd5d6375860072de44e1e9740","cross_cats_sorted":["cs.AI","cs.CL"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-02-28T18:23:49Z","title_canon_sha256":"92c76a4ca142260c092afea071ddc39d118d5471cf5ea303165af94fa5f97dd0"},"schema_version":"1.0","source":{"id":"2402.18540","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2402.18540","created_at":"2026-07-05T10:02:02Z"},{"alias_kind":"arxiv_version","alias_value":"2402.18540v2","created_at":"2026-07-05T10:02:02Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2402.18540","created_at":"2026-07-05T10:02:02Z"},{"alias_kind":"pith_short_12","alias_value":"GGXWKXV6A6KQ","created_at":"2026-07-05T10:02:02Z"},{"alias_kind":"pith_short_16","alias_value":"GGXWKXV6A6KQ5MME","created_at":"2026-07-05T10:02:02Z"},{"alias_kind":"pith_short_8","alias_value":"GGXWKXV6","created_at":"2026-07-05T10:02:02Z"}],"graph_snapshots":[{"event_id":"sha256:c0b20d371c6afd230abcbbc2c555e2988b12c852f6b10c19a2665d83dc0d87da","target":"graph","created_at":"2026-07-05T10:02:02Z","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/2402.18540/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Public LLMs such as the Llama 2-Chat underwent alignment training and were considered safe. Recently Qi et al. [2024] reported that even benign fine-tuning on seemingly safe datasets can give rise to unsafe behaviors in the models. The current paper is about methods and best practices to mitigate such loss of alignment. We focus on the setting where a public model is fine-tuned before serving users for specific usage, where the model should improve on the downstream task while maintaining alignment. Through extensive experiments on several chat models (Meta's Llama 2-Chat, Mistral AI's Mistral","authors_text":"Anirudh Goyal, Dingli Yu, Haoyu Zhao, Kaifeng Lyu, Sanjeev Arora, Xinran Gu","cross_cats":["cs.AI","cs.CL"],"headline":"","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-02-28T18:23:49Z","title":"Keeping LLMs Aligned After Fine-tuning: The Crucial Role of Prompt Templates"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2402.18540","kind":"arxiv","version":2},"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:f2c8ff06f7d2cd733fc800342212e504b01131cc417e52177c38a066b0d71458","target":"record","created_at":"2026-07-05T10:02:02Z","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":"711a9bd39d6e66517ee976a6d6309caba66c335fd5d6375860072de44e1e9740","cross_cats_sorted":["cs.AI","cs.CL"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-02-28T18:23:49Z","title_canon_sha256":"92c76a4ca142260c092afea071ddc39d118d5471cf5ea303165af94fa5f97dd0"},"schema_version":"1.0","source":{"id":"2402.18540","kind":"arxiv","version":2}},"canonical_sha256":"31af655ebe07950eb1845e75a560b763df8d7cda9b8c2ae03726f9e843d0cb16","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"31af655ebe07950eb1845e75a560b763df8d7cda9b8c2ae03726f9e843d0cb16","first_computed_at":"2026-07-05T10:02:02.302594Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T10:02:02.302594Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"qCzUXp8bnZNi3n6jhxeU1yG6w1pWNqNpWntsBZos2qy/l+3rDdjdouh4m9Y0oAezuf7WlLCf14G9hd5tmqWGDA==","signature_status":"signed_v1","signed_at":"2026-07-05T10:02:02.303013Z","signed_message":"canonical_sha256_bytes"},"source_id":"2402.18540","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:f2c8ff06f7d2cd733fc800342212e504b01131cc417e52177c38a066b0d71458","sha256:c0b20d371c6afd230abcbbc2c555e2988b12c852f6b10c19a2665d83dc0d87da"],"state_sha256":"3d0a93e17b01eca30c044779a14f8a0a81cc52921f2e5f67b605c66c4c296d54"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"DxfcGxsJPVFBUQ3kGFfHJZQYbTKZPlRb8IsE2sZpXg/2oEk945DZWB8WhfrTWyvw2P2NgaTqJRzHhMn+Q0yJDg==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-08T13:56:09.151990Z","bundle_sha256":"08ea4aa9530a7f6012a8e53d8a88dfe32c5341cb907bc4ee97e7c766f410ca47"}}