{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:OKPXZKYJOA6SCRDCIWVR2YK6PG","short_pith_number":"pith:OKPXZKYJ","schema_version":"1.0","canonical_sha256":"729f7cab09703d21446245ab1d615e7980af420d76d542dd43b2d377ebbc35e5","source":{"kind":"arxiv","id":"2510.03485","version":2},"attestation_state":"computed","paper":{"title":"Learning Efficient Guardrails for Compliance","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Muhao Chen, Peng Qi, Wenjie Jacky Mo, Xiaofei Wen, Yanan Xie","submitted_at":"2025-10-03T20:03:19Z","abstract_excerpt":"Autonomous web agents are increasingly deployed for long-horizon tasks, yet their ability to adhere to real-world policies remains critically underexplored compared to standard safety objectives. To address this gap, we introduce PolicyGuardBench, a benchmark of 60k policy-trajectory pairs designed to evaluate compliance through both full-trajectory and novel prefix-based violation detection tasks. Using this dataset, we train PolicyGuard, a lightweight guardrail model that achieves strong detection accuracy while maintaining high inference efficiency. Notably, our model demonstrates robust ge"},"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":"2510.03485","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by-sa/4.0/","primary_cat":"cs.AI","submitted_at":"2025-10-03T20:03:19Z","cross_cats_sorted":[],"title_canon_sha256":"5f89b067d5dd3def0c4993f6684fb7310002f8b5690f913b336495c2c3e5e1f5","abstract_canon_sha256":"b7da4c768dc3d85ffdde7f310843d0ce3ef70f3ab7ad07b33b03239867cd428a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T01:04:59.202635Z","signature_b64":"aMsy9No+GsLeDPWvID/lcyFgzejsqnpHt25usGtEi8PVNYqcPI/ZajY/QC1ua6OsmG2lnIVtKZzuFAftwSIsDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"729f7cab09703d21446245ab1d615e7980af420d76d542dd43b2d377ebbc35e5","last_reissued_at":"2026-05-20T01:04:59.201852Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T01:04:59.201852Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Learning Efficient Guardrails for Compliance","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Muhao Chen, Peng Qi, Wenjie Jacky Mo, Xiaofei Wen, Yanan Xie","submitted_at":"2025-10-03T20:03:19Z","abstract_excerpt":"Autonomous web agents are increasingly deployed for long-horizon tasks, yet their ability to adhere to real-world policies remains critically underexplored compared to standard safety objectives. To address this gap, we introduce PolicyGuardBench, a benchmark of 60k policy-trajectory pairs designed to evaluate compliance through both full-trajectory and novel prefix-based violation detection tasks. Using this dataset, we train PolicyGuard, a lightweight guardrail model that achieves strong detection accuracy while maintaining high inference efficiency. Notably, our model demonstrates robust ge"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2510.03485","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/2510.03485/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":"2510.03485","created_at":"2026-05-20T01:04:59.201968+00:00"},{"alias_kind":"arxiv_version","alias_value":"2510.03485v2","created_at":"2026-05-20T01:04:59.201968+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2510.03485","created_at":"2026-05-20T01:04:59.201968+00:00"},{"alias_kind":"pith_short_12","alias_value":"OKPXZKYJOA6S","created_at":"2026-05-20T01:04:59.201968+00:00"},{"alias_kind":"pith_short_16","alias_value":"OKPXZKYJOA6SCRDC","created_at":"2026-05-20T01:04:59.201968+00:00"},{"alias_kind":"pith_short_8","alias_value":"OKPXZKYJ","created_at":"2026-05-20T01:04:59.201968+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2605.17329","citing_title":"LPG: Balancing Efficiency and Policy Reasoning in Latent Policy Guardrails","ref_index":33,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/OKPXZKYJOA6SCRDCIWVR2YK6PG","json":"https://pith.science/pith/OKPXZKYJOA6SCRDCIWVR2YK6PG.json","graph_json":"https://pith.science/api/pith-number/OKPXZKYJOA6SCRDCIWVR2YK6PG/graph.json","events_json":"https://pith.science/api/pith-number/OKPXZKYJOA6SCRDCIWVR2YK6PG/events.json","paper":"https://pith.science/paper/OKPXZKYJ"},"agent_actions":{"view_html":"https://pith.science/pith/OKPXZKYJOA6SCRDCIWVR2YK6PG","download_json":"https://pith.science/pith/OKPXZKYJOA6SCRDCIWVR2YK6PG.json","view_paper":"https://pith.science/paper/OKPXZKYJ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2510.03485&json=true","fetch_graph":"https://pith.science/api/pith-number/OKPXZKYJOA6SCRDCIWVR2YK6PG/graph.json","fetch_events":"https://pith.science/api/pith-number/OKPXZKYJOA6SCRDCIWVR2YK6PG/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/OKPXZKYJOA6SCRDCIWVR2YK6PG/action/timestamp_anchor","attest_storage":"https://pith.science/pith/OKPXZKYJOA6SCRDCIWVR2YK6PG/action/storage_attestation","attest_author":"https://pith.science/pith/OKPXZKYJOA6SCRDCIWVR2YK6PG/action/author_attestation","sign_citation":"https://pith.science/pith/OKPXZKYJOA6SCRDCIWVR2YK6PG/action/citation_signature","submit_replication":"https://pith.science/pith/OKPXZKYJOA6SCRDCIWVR2YK6PG/action/replication_record"}},"created_at":"2026-05-20T01:04:59.201968+00:00","updated_at":"2026-05-20T01:04:59.201968+00:00"}