{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:VCTQAC3HZCA7FFRSOZFA2F46EB","short_pith_number":"pith:VCTQAC3H","schema_version":"1.0","canonical_sha256":"a8a7000b67c881f29632764a0d179e2055c5197902469593393e87cb58e8bdc6","source":{"kind":"arxiv","id":"2605.31073","version":1},"attestation_state":"computed","paper":{"title":"ConsisGuard: Aligning Safety Deliberation with Policy Enforcement in LLM Guardrails","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Bingyu Zhu, Hui Xue, Jungang Lou, Kui Ren, Longtao Huang, Ningyu Zhang, Yan Wang, Yuefeng Chen, Zeyu Yang, Zhen Bi, Zhixuan Chu, Zihao Xue","submitted_at":"2026-05-29T09:42:08Z","abstract_excerpt":"Reasoning-based LLM guardrails improve safety moderation by generating explicit rationales before issuing final decisions. However, their rationales do not always lead to faithful enforcement: a model may recognize a harmful intent in its reasoning but still predict a safe label, or issue an unsafe decision without policy-grounded justification. We identify this safety-critical failure mode as the deliberation-to-enforcement gap. Unlike general chain-of-thought faithfulness, guardrail reliability requires policy execution consistency: the generated reasoning should be grounded in the safety po"},"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":"2605.31073","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-05-29T09:42:08Z","cross_cats_sorted":[],"title_canon_sha256":"1ac1ab4e5752415e59c92cc880bea9c4961830302efa836ffed8d69a9c9ab080","abstract_canon_sha256":"5dc581694927bec36b0b5796fcaeac6f1c209faadb0fd2525595422f55ee6c53"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-01T01:03:34.640003Z","signature_b64":"x3NFxCJPRIHrCjry9hn1tYfrfJY7F4i+xHnIVpjJ6Saz1J0sZBqFFzynDK7xrCGEkGpt3ger3VP9ht5dSHC4Cg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a8a7000b67c881f29632764a0d179e2055c5197902469593393e87cb58e8bdc6","last_reissued_at":"2026-06-01T01:03:34.639425Z","signature_status":"signed_v1","first_computed_at":"2026-06-01T01:03:34.639425Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"ConsisGuard: Aligning Safety Deliberation with Policy Enforcement in LLM Guardrails","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Bingyu Zhu, Hui Xue, Jungang Lou, Kui Ren, Longtao Huang, Ningyu Zhang, Yan Wang, Yuefeng Chen, Zeyu Yang, Zhen Bi, Zhixuan Chu, Zihao Xue","submitted_at":"2026-05-29T09:42:08Z","abstract_excerpt":"Reasoning-based LLM guardrails improve safety moderation by generating explicit rationales before issuing final decisions. However, their rationales do not always lead to faithful enforcement: a model may recognize a harmful intent in its reasoning but still predict a safe label, or issue an unsafe decision without policy-grounded justification. We identify this safety-critical failure mode as the deliberation-to-enforcement gap. Unlike general chain-of-thought faithfulness, guardrail reliability requires policy execution consistency: the generated reasoning should be grounded in the safety po"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.31073","kind":"arxiv","version":1},"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/2605.31073/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":"2605.31073","created_at":"2026-06-01T01:03:34.639521+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.31073v1","created_at":"2026-06-01T01:03:34.639521+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.31073","created_at":"2026-06-01T01:03:34.639521+00:00"},{"alias_kind":"pith_short_12","alias_value":"VCTQAC3HZCA7","created_at":"2026-06-01T01:03:34.639521+00:00"},{"alias_kind":"pith_short_16","alias_value":"VCTQAC3HZCA7FFRS","created_at":"2026-06-01T01:03:34.639521+00:00"},{"alias_kind":"pith_short_8","alias_value":"VCTQAC3H","created_at":"2026-06-01T01:03:34.639521+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/VCTQAC3HZCA7FFRSOZFA2F46EB","json":"https://pith.science/pith/VCTQAC3HZCA7FFRSOZFA2F46EB.json","graph_json":"https://pith.science/api/pith-number/VCTQAC3HZCA7FFRSOZFA2F46EB/graph.json","events_json":"https://pith.science/api/pith-number/VCTQAC3HZCA7FFRSOZFA2F46EB/events.json","paper":"https://pith.science/paper/VCTQAC3H"},"agent_actions":{"view_html":"https://pith.science/pith/VCTQAC3HZCA7FFRSOZFA2F46EB","download_json":"https://pith.science/pith/VCTQAC3HZCA7FFRSOZFA2F46EB.json","view_paper":"https://pith.science/paper/VCTQAC3H","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.31073&json=true","fetch_graph":"https://pith.science/api/pith-number/VCTQAC3HZCA7FFRSOZFA2F46EB/graph.json","fetch_events":"https://pith.science/api/pith-number/VCTQAC3HZCA7FFRSOZFA2F46EB/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/VCTQAC3HZCA7FFRSOZFA2F46EB/action/timestamp_anchor","attest_storage":"https://pith.science/pith/VCTQAC3HZCA7FFRSOZFA2F46EB/action/storage_attestation","attest_author":"https://pith.science/pith/VCTQAC3HZCA7FFRSOZFA2F46EB/action/author_attestation","sign_citation":"https://pith.science/pith/VCTQAC3HZCA7FFRSOZFA2F46EB/action/citation_signature","submit_replication":"https://pith.science/pith/VCTQAC3HZCA7FFRSOZFA2F46EB/action/replication_record"}},"created_at":"2026-06-01T01:03:34.639521+00:00","updated_at":"2026-06-01T01:03:34.639521+00:00"}