{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2024:2P2SDAYE5WMPKCBSAMHDKEQODO","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":"4efe0c27c6c02921251076ca3800f38e0827c9831b7eb72e2ba8f181997c141e","cross_cats_sorted":["cs.AI","cs.CL"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-01-31T17:28:24Z","title_canon_sha256":"72d132d4314cbb87869a5a09f63bc2abdcf3844b51230ca308d47b7a67fbd15a"},"schema_version":"1.0","source":{"id":"2401.18018","kind":"arxiv","version":4}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2401.18018","created_at":"2026-07-05T08:26:18Z"},{"alias_kind":"arxiv_version","alias_value":"2401.18018v4","created_at":"2026-07-05T08:26:18Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2401.18018","created_at":"2026-07-05T08:26:18Z"},{"alias_kind":"pith_short_12","alias_value":"2P2SDAYE5WMP","created_at":"2026-07-05T08:26:18Z"},{"alias_kind":"pith_short_16","alias_value":"2P2SDAYE5WMPKCBS","created_at":"2026-07-05T08:26:18Z"},{"alias_kind":"pith_short_8","alias_value":"2P2SDAYE","created_at":"2026-07-05T08:26:18Z"}],"graph_snapshots":[{"event_id":"sha256:8c3dcf6ce7f77db297d288a1542eafee7b11c897ab4171f7cbebfb7068a4d461","target":"graph","created_at":"2026-07-05T08:26:18Z","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/2401.18018/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Prepending model inputs with safety prompts is a common practice for safeguarding large language models (LLMs) against queries with harmful intents. However, the underlying working mechanisms of safety prompts have not been unraveled yet, restricting the possibility of automatically optimizing them to improve LLM safety. In this work, we investigate how LLMs' behavior (i.e., complying with or refusing user queries) is affected by safety prompts from the perspective of model representation. We find that in the representation space, the input queries are typically moved by safety prompts in a \"h","authors_text":"Chujie Zheng, Fandong Meng, Fan Yin, Hao Zhou, Jie Zhou, Kai-Wei Chang, Minlie Huang, Nanyun Peng","cross_cats":["cs.AI","cs.CL"],"headline":"","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-01-31T17:28:24Z","title":"On Prompt-Driven Safeguarding for Large Language Models"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2401.18018","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:be0454a1d83be31d5d3bf81d32e41d9470335762c12d2e6edd97335f75f61b0d","target":"record","created_at":"2026-07-05T08:26:18Z","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":"4efe0c27c6c02921251076ca3800f38e0827c9831b7eb72e2ba8f181997c141e","cross_cats_sorted":["cs.AI","cs.CL"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-01-31T17:28:24Z","title_canon_sha256":"72d132d4314cbb87869a5a09f63bc2abdcf3844b51230ca308d47b7a67fbd15a"},"schema_version":"1.0","source":{"id":"2401.18018","kind":"arxiv","version":4}},"canonical_sha256":"d3f5218304ed98f50832030e35120e1b825ad08d19e5adb88cfa062134cc8f60","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"d3f5218304ed98f50832030e35120e1b825ad08d19e5adb88cfa062134cc8f60","first_computed_at":"2026-07-05T08:26:18.394145Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T08:26:18.394145Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"a5vM64Old6lLcAT69uJqmW7aQ77/P/D+JNfNaCDYOkrsamedMwEw1eF7ENtmMWPNNbNqQ0p2JphLdSh55TnWCg==","signature_status":"signed_v1","signed_at":"2026-07-05T08:26:18.394643Z","signed_message":"canonical_sha256_bytes"},"source_id":"2401.18018","source_kind":"arxiv","source_version":4}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:be0454a1d83be31d5d3bf81d32e41d9470335762c12d2e6edd97335f75f61b0d","sha256:8c3dcf6ce7f77db297d288a1542eafee7b11c897ab4171f7cbebfb7068a4d461"],"state_sha256":"ab8e545be6763b856e4c2e1eff9cb809c306388811dbd21651c445d01dd7c861"}