{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:YZHGGI3WXG364LWKXMKD7QJIVN","short_pith_number":"pith:YZHGGI3W","schema_version":"1.0","canonical_sha256":"c64e632376b9b7ee2ecabb143fc128ab42e6b83bbe48e1960841ba8e63772a2b","source":{"kind":"arxiv","id":"2502.15594","version":2},"attestation_state":"computed","paper":{"title":"SafeInt: Shielding Large Language Models from Jailbreak Attacks via Safety-Aware Representation Intervention","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Chen Chen, Chunyan Hou, Jiaqi Wu, Xiaojie Yuan","submitted_at":"2025-02-21T17:12:35Z","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"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2502.15594","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-02-21T17:12:35Z","cross_cats_sorted":[],"title_canon_sha256":"d698d593f96063f36c825ffcd56705d7876a117e3ffe151086d18d36dbc0eb3a","abstract_canon_sha256":"3f9849d9ee1f41041708e8b9d3e306d976bb251e85162e82b48c25548a36efd1"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:08:17.047483Z","signature_b64":"Oe0woZ2O/L3zCdEEpa5Wou9EDFD6EXVKDFyQWw7PX6BxqaE6/OoSKwOf9WHM/0VZG+NkPfG7xp0ESIbep/TRBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"c64e632376b9b7ee2ecabb143fc128ab42e6b83bbe48e1960841ba8e63772a2b","last_reissued_at":"2026-07-05T11:08:17.046929Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:08:17.046929Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"SafeInt: Shielding Large Language Models from Jailbreak Attacks via Safety-Aware Representation Intervention","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Chen Chen, Chunyan Hou, Jiaqi Wu, Xiaojie Yuan","submitted_at":"2025-02-21T17:12:35Z","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"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.15594","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/2502.15594/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"},"aliases":[{"alias_kind":"arxiv","alias_value":"2502.15594","created_at":"2026-07-05T11:08:17.046992+00:00"},{"alias_kind":"arxiv_version","alias_value":"2502.15594v2","created_at":"2026-07-05T11:08:17.046992+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.15594","created_at":"2026-07-05T11:08:17.046992+00:00"},{"alias_kind":"pith_short_12","alias_value":"YZHGGI3WXG36","created_at":"2026-07-05T11:08:17.046992+00:00"},{"alias_kind":"pith_short_16","alias_value":"YZHGGI3WXG364LWK","created_at":"2026-07-05T11:08:17.046992+00:00"},{"alias_kind":"pith_short_8","alias_value":"YZHGGI3W","created_at":"2026-07-05T11:08:17.046992+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/YZHGGI3WXG364LWKXMKD7QJIVN","json":"https://pith.science/pith/YZHGGI3WXG364LWKXMKD7QJIVN.json","graph_json":"https://pith.science/api/pith-number/YZHGGI3WXG364LWKXMKD7QJIVN/graph.json","events_json":"https://pith.science/api/pith-number/YZHGGI3WXG364LWKXMKD7QJIVN/events.json","paper":"https://pith.science/paper/YZHGGI3W"},"agent_actions":{"view_html":"https://pith.science/pith/YZHGGI3WXG364LWKXMKD7QJIVN","download_json":"https://pith.science/pith/YZHGGI3WXG364LWKXMKD7QJIVN.json","view_paper":"https://pith.science/paper/YZHGGI3W","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2502.15594&json=true","fetch_graph":"https://pith.science/api/pith-number/YZHGGI3WXG364LWKXMKD7QJIVN/graph.json","fetch_events":"https://pith.science/api/pith-number/YZHGGI3WXG364LWKXMKD7QJIVN/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/YZHGGI3WXG364LWKXMKD7QJIVN/action/timestamp_anchor","attest_storage":"https://pith.science/pith/YZHGGI3WXG364LWKXMKD7QJIVN/action/storage_attestation","attest_author":"https://pith.science/pith/YZHGGI3WXG364LWKXMKD7QJIVN/action/author_attestation","sign_citation":"https://pith.science/pith/YZHGGI3WXG364LWKXMKD7QJIVN/action/citation_signature","submit_replication":"https://pith.science/pith/YZHGGI3WXG364LWKXMKD7QJIVN/action/replication_record"}},"created_at":"2026-07-05T11:08:17.046992+00:00","updated_at":"2026-07-05T11:08:17.046992+00:00"}