{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2025:YZHGGI3WXG364LWKXMKD7QJIVN","short_pith_number":"pith:YZHGGI3W","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"},"canonical_sha256":"c64e632376b9b7ee2ecabb143fc128ab42e6b83bbe48e1960841ba8e63772a2b","source":{"kind":"arxiv","id":"2502.15594","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"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2025:YZHGGI3WXG364LWKXMKD7QJIVN","target":"record","payload":{"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"},"canonical_sha256":"c64e632376b9b7ee2ecabb143fc128ab42e6b83bbe48e1960841ba8e63772a2b","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"},"source_kind":"arxiv","source_id":"2502.15594","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":"2026-07-05T11:08:17Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"6gfGPbwMe8kYlIpzzQQP6K67C5nF5KU0MeROSgFWBDR/Wyq34u14/JEhLgUobqxoApW/9Kn0zxXQ07UqiGNgCw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-03T15:56:12.094799Z"},"content_sha256":"540ab1f4852822d0757a795e504ea44031229582f3d6fcb90eacf5b4219e8669","schema_version":"1.0","event_id":"sha256:540ab1f4852822d0757a795e504ea44031229582f3d6fcb90eacf5b4219e8669"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2025:YZHGGI3WXG364LWKXMKD7QJIVN","target":"graph","payload":{"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"},"verdict_id":null},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T11:08:17Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"jjtb5m6wERjlzjrAklkOqU+D1afHCC+8QyCaD14vM9X0xir/UaqjKyaixCXaonOofiSOJoXOEM9leVVGYVN5CQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-03T15:56:12.095732Z"},"content_sha256":"4d6a772e520dfa02b46f9a27e59777edc617a4cd3c3979c2f2d5c2e4bf112500","schema_version":"1.0","event_id":"sha256:4d6a772e520dfa02b46f9a27e59777edc617a4cd3c3979c2f2d5c2e4bf112500"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/YZHGGI3WXG364LWKXMKD7QJIVN/bundle.json","state_url":"https://pith.science/pith/YZHGGI3WXG364LWKXMKD7QJIVN/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/YZHGGI3WXG364LWKXMKD7QJIVN/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":"2026-08-03T15:56:12Z","links":{"resolver":"https://pith.science/pith/YZHGGI3WXG364LWKXMKD7QJIVN","bundle":"https://pith.science/pith/YZHGGI3WXG364LWKXMKD7QJIVN/bundle.json","state":"https://pith.science/pith/YZHGGI3WXG364LWKXMKD7QJIVN/state.json","well_known_bundle":"https://pith.science/.well-known/pith/YZHGGI3WXG364LWKXMKD7QJIVN/bundle.json"},"state":{"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"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"Q95ol/MTIPewMhM6T5dXj/kgbMYot18HC3a177N21/Ov8+djlSRdYAIHrEYhHVv0XUII8lvuBqqSoEG6dxn3Cg==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-03T15:56:12.105073Z","bundle_sha256":"3ec108c84206c5e2331615a15bfdf1fece9eaf6ce1adc7f1d14b0c260bf8a99f"}}