{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2025:YZHGGI3WXG364LWKXMKD7QJIVN","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":"3f9849d9ee1f41041708e8b9d3e306d976bb251e85162e82b48c25548a36efd1","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-02-21T17:12:35Z","title_canon_sha256":"d698d593f96063f36c825ffcd56705d7876a117e3ffe151086d18d36dbc0eb3a"},"schema_version":"1.0","source":{"id":"2502.15594","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2502.15594","created_at":"2026-07-05T11:08:17Z"},{"alias_kind":"arxiv_version","alias_value":"2502.15594v2","created_at":"2026-07-05T11:08:17Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.15594","created_at":"2026-07-05T11:08:17Z"},{"alias_kind":"pith_short_12","alias_value":"YZHGGI3WXG36","created_at":"2026-07-05T11:08:17Z"},{"alias_kind":"pith_short_16","alias_value":"YZHGGI3WXG364LWK","created_at":"2026-07-05T11:08:17Z"},{"alias_kind":"pith_short_8","alias_value":"YZHGGI3W","created_at":"2026-07-05T11:08:17Z"}],"graph_snapshots":[{"event_id":"sha256:4d6a772e520dfa02b46f9a27e59777edc617a4cd3c3979c2f2d5c2e4bf112500","target":"graph","created_at":"2026-07-05T11:08:17Z","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/2502.15594/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"With the widespread real-world deployment of large language models (LLMs), ensuring their behavior complies with safety standards has become crucial. Jailbreak attacks exploit vulnerabilities in LLMs to induce undesirable behavior, posing a significant threat to LLM safety. Previous defenses often fail to achieve both effectiveness and efficiency simultaneously. Defenses from a representation perspective offer new insights, but existing interventions cannot dynamically adjust representations based on the harmfulness of the queries. To address this limitation, we propose SafeIntervention (SafeI","authors_text":"Chen Chen, Chunyan Hou, Jiaqi Wu, Xiaojie Yuan","cross_cats":[],"headline":"","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-02-21T17:12:35Z","title":"SafeInt: Shielding Large Language Models from Jailbreak Attacks via Safety-Aware Representation Intervention"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.15594","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":"sha256:540ab1f4852822d0757a795e504ea44031229582f3d6fcb90eacf5b4219e8669","target":"record","created_at":"2026-07-05T11:08:17Z","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":"3f9849d9ee1f41041708e8b9d3e306d976bb251e85162e82b48c25548a36efd1","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-02-21T17:12:35Z","title_canon_sha256":"d698d593f96063f36c825ffcd56705d7876a117e3ffe151086d18d36dbc0eb3a"},"schema_version":"1.0","source":{"id":"2502.15594","kind":"arxiv","version":2}},"canonical_sha256":"c64e632376b9b7ee2ecabb143fc128ab42e6b83bbe48e1960841ba8e63772a2b","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"c64e632376b9b7ee2ecabb143fc128ab42e6b83bbe48e1960841ba8e63772a2b","first_computed_at":"2026-07-05T11:08:17.046929Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T11:08:17.046929Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"Oe0woZ2O/L3zCdEEpa5Wou9EDFD6EXVKDFyQWw7PX6BxqaE6/OoSKwOf9WHM/0VZG+NkPfG7xp0ESIbep/TRBg==","signature_status":"signed_v1","signed_at":"2026-07-05T11:08:17.047483Z","signed_message":"canonical_sha256_bytes"},"source_id":"2502.15594","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:540ab1f4852822d0757a795e504ea44031229582f3d6fcb90eacf5b4219e8669","sha256:4d6a772e520dfa02b46f9a27e59777edc617a4cd3c3979c2f2d5c2e4bf112500"],"state_sha256":"3497db0f2c30190e87fe7d8d9768d25c0fb899263d88c3992493b49ee412c681"}