{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2024:JGMOODPUKUI6WCX456WSHCGV2N","short_pith_number":"pith:JGMOODPU","canonical_record":{"source":{"id":"2402.11889","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-02-19T06:58:42Z","cross_cats_sorted":[],"title_canon_sha256":"fcaabf66dd00f0337a1f04fd5cf86edc6bd822c1e61a98fe1f197c73bb103ae1","abstract_canon_sha256":"2f4caa68adf2dd6427658ae06a2923042942979080fc092403a1abcdf18eac61"},"schema_version":"1.0"},"canonical_sha256":"4998e70df45511eb0afcefad2388d5d367bdeed124d459f96c089d4a1d38f8cd","source":{"kind":"arxiv","id":"2402.11889","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2402.11889","created_at":"2026-07-05T08:32:37Z"},{"alias_kind":"arxiv_version","alias_value":"2402.11889v2","created_at":"2026-07-05T08:32:37Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2402.11889","created_at":"2026-07-05T08:32:37Z"},{"alias_kind":"pith_short_12","alias_value":"JGMOODPUKUI6","created_at":"2026-07-05T08:32:37Z"},{"alias_kind":"pith_short_16","alias_value":"JGMOODPUKUI6WCX4","created_at":"2026-07-05T08:32:37Z"},{"alias_kind":"pith_short_8","alias_value":"JGMOODPU","created_at":"2026-07-05T08:32:37Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2024:JGMOODPUKUI6WCX456WSHCGV2N","target":"record","payload":{"canonical_record":{"source":{"id":"2402.11889","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-02-19T06:58:42Z","cross_cats_sorted":[],"title_canon_sha256":"fcaabf66dd00f0337a1f04fd5cf86edc6bd822c1e61a98fe1f197c73bb103ae1","abstract_canon_sha256":"2f4caa68adf2dd6427658ae06a2923042942979080fc092403a1abcdf18eac61"},"schema_version":"1.0"},"canonical_sha256":"4998e70df45511eb0afcefad2388d5d367bdeed124d459f96c089d4a1d38f8cd","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:32:37.007341Z","signature_b64":"TWj88rJ/muAgNLPpzNUoCVKTjKgnk5Lqbi3PtScozXMzpCCLx+xegrC8o/YO3joq30zB9qREVWJKcxU4TLUhDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"4998e70df45511eb0afcefad2388d5d367bdeed124d459f96c089d4a1d38f8cd","last_reissued_at":"2026-07-05T08:32:37.006815Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:32:37.006815Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2402.11889","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":"2026-07-05T08:32:37Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"PPT2NbJLTAoxqInZQNKuQAvR6mvUykoPIx+hGk7RAUtRg7wQvtzPPVeFEJwcBKhFZCbHdqd7w3lIBnlBILACAw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-18T00:32:37.281813Z"},"content_sha256":"3dc36b7e2b3fa16862b4b52f5ffcf309162adee8739df48d61f9ba09ca4c0083","schema_version":"1.0","event_id":"sha256:3dc36b7e2b3fa16862b4b52f5ffcf309162adee8739df48d61f9ba09ca4c0083"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2024:JGMOODPUKUI6WCX456WSHCGV2N","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"ROSE Doesn't Do That: Boosting the Safety of Instruction-Tuned Large Language Models with Reverse Prompt Contrastive Decoding","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Bo Du, Dacheng Tao, Juhua Liu, Liang Ding, Qihuang Zhong","submitted_at":"2024-02-19T06:58:42Z","abstract_excerpt":"With the development of instruction-tuned large language models (LLMs), improving the safety of LLMs has become more critical. However, the current approaches for aligning the LLMs output with expected safety usually require substantial training efforts, e.g., high-quality safety data and expensive computational resources, which are costly and inefficient. To this end, we present reverse prompt contrastive decoding (ROSE), a simple-yet-effective method to directly boost the safety of existing instruction-tuned LLMs without any additional training. The principle of ROSE is to improve the probab"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2402.11889","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/2402.11889/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":"2026-07-05T08:32:37Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"C6wYLIKLS3gWAaK7jB67Rxqninat26TzuGqqWxWK9TTSDt4UNsF1muMuqOVk216hvz0wHrtl5MGB0rXD4qgyDQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-18T00:32:37.282404Z"},"content_sha256":"62594a8a8c190d3d6dae50bce46c5ba55fb2b96e316e58d1d989d06751983df8","schema_version":"1.0","event_id":"sha256:62594a8a8c190d3d6dae50bce46c5ba55fb2b96e316e58d1d989d06751983df8"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/JGMOODPUKUI6WCX456WSHCGV2N/bundle.json","state_url":"https://pith.science/pith/JGMOODPUKUI6WCX456WSHCGV2N/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/JGMOODPUKUI6WCX456WSHCGV2N/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":"2026-08-18T00:32:37Z","links":{"resolver":"https://pith.science/pith/JGMOODPUKUI6WCX456WSHCGV2N","bundle":"https://pith.science/pith/JGMOODPUKUI6WCX456WSHCGV2N/bundle.json","state":"https://pith.science/pith/JGMOODPUKUI6WCX456WSHCGV2N/state.json","well_known_bundle":"https://pith.science/.well-known/pith/JGMOODPUKUI6WCX456WSHCGV2N/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2024:JGMOODPUKUI6WCX456WSHCGV2N","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":"2f4caa68adf2dd6427658ae06a2923042942979080fc092403a1abcdf18eac61","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-02-19T06:58:42Z","title_canon_sha256":"fcaabf66dd00f0337a1f04fd5cf86edc6bd822c1e61a98fe1f197c73bb103ae1"},"schema_version":"1.0","source":{"id":"2402.11889","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2402.11889","created_at":"2026-07-05T08:32:37Z"},{"alias_kind":"arxiv_version","alias_value":"2402.11889v2","created_at":"2026-07-05T08:32:37Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2402.11889","created_at":"2026-07-05T08:32:37Z"},{"alias_kind":"pith_short_12","alias_value":"JGMOODPUKUI6","created_at":"2026-07-05T08:32:37Z"},{"alias_kind":"pith_short_16","alias_value":"JGMOODPUKUI6WCX4","created_at":"2026-07-05T08:32:37Z"},{"alias_kind":"pith_short_8","alias_value":"JGMOODPU","created_at":"2026-07-05T08:32:37Z"}],"graph_snapshots":[{"event_id":"sha256:62594a8a8c190d3d6dae50bce46c5ba55fb2b96e316e58d1d989d06751983df8","target":"graph","created_at":"2026-07-05T08:32:37Z","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/2402.11889/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"With the development of instruction-tuned large language models (LLMs), improving the safety of LLMs has become more critical. However, the current approaches for aligning the LLMs output with expected safety usually require substantial training efforts, e.g., high-quality safety data and expensive computational resources, which are costly and inefficient. To this end, we present reverse prompt contrastive decoding (ROSE), a simple-yet-effective method to directly boost the safety of existing instruction-tuned LLMs without any additional training. The principle of ROSE is to improve the probab","authors_text":"Bo Du, Dacheng Tao, Juhua Liu, Liang Ding, Qihuang Zhong","cross_cats":[],"headline":"","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-02-19T06:58:42Z","title":"ROSE Doesn't Do That: Boosting the Safety of Instruction-Tuned Large Language Models with Reverse Prompt Contrastive Decoding"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2402.11889","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":"sha256:3dc36b7e2b3fa16862b4b52f5ffcf309162adee8739df48d61f9ba09ca4c0083","target":"record","created_at":"2026-07-05T08:32:37Z","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":"2f4caa68adf2dd6427658ae06a2923042942979080fc092403a1abcdf18eac61","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-02-19T06:58:42Z","title_canon_sha256":"fcaabf66dd00f0337a1f04fd5cf86edc6bd822c1e61a98fe1f197c73bb103ae1"},"schema_version":"1.0","source":{"id":"2402.11889","kind":"arxiv","version":2}},"canonical_sha256":"4998e70df45511eb0afcefad2388d5d367bdeed124d459f96c089d4a1d38f8cd","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"4998e70df45511eb0afcefad2388d5d367bdeed124d459f96c089d4a1d38f8cd","first_computed_at":"2026-07-05T08:32:37.006815Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T08:32:37.006815Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"TWj88rJ/muAgNLPpzNUoCVKTjKgnk5Lqbi3PtScozXMzpCCLx+xegrC8o/YO3joq30zB9qREVWJKcxU4TLUhDw==","signature_status":"signed_v1","signed_at":"2026-07-05T08:32:37.007341Z","signed_message":"canonical_sha256_bytes"},"source_id":"2402.11889","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:3dc36b7e2b3fa16862b4b52f5ffcf309162adee8739df48d61f9ba09ca4c0083","sha256:62594a8a8c190d3d6dae50bce46c5ba55fb2b96e316e58d1d989d06751983df8"],"state_sha256":"4b4430b1c2fc1c54ef033cfbc9eea290bcfc2e5cf9b59062ea77b1dc459bb5e8"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"8G+i/9MteIj77drWrcmTODT5EIweC0yOipRHX8jFWYucGlaW7FqcH2Ylw/xG/7xSZIsGYuQAqdLDwBz47Ee3Ag==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-18T00:32:37.285966Z","bundle_sha256":"ad22a534e79bc138adf7d91551c2d2a30f52b483d5502da53746bd199bd78ef5"}}