{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:QEBEAEEO3REGVPIBHYPLFO2IAF","short_pith_number":"pith:QEBEAEEO","schema_version":"1.0","canonical_sha256":"810240108edc486abd013e1eb2bb48014571dbdd3e6a1b128b770e1ee635a1a9","source":{"kind":"arxiv","id":"2603.00829","version":2},"attestation_state":"computed","paper":{"title":"Constitutional Black-Box Monitoring for Scheming in LLM Agents","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"James Peters-Gill, Marius Hobbhahn, Rich Barton-Cooper, Simon Storf","submitted_at":"2026-02-28T22:31:32Z","abstract_excerpt":"Safe deployment of Large Language Model (LLM) agents in autonomous settings requires reliable oversight mechanisms. A central challenge is detecting scheming, where agents covertly pursue misaligned goals. One approach to mitigating such risks is LLM-based monitoring: using language models to examine agent behaviors for suspicious actions. We study constitutional black-box monitors: prompted classifiers that detect scheming using only externally observable inputs and outputs, optimized on synthetic data generated from natural-language behavior specifications. We introduce two pipelines for gen"},"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":"2603.00829","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-02-28T22:31:32Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"f6412d3f5f7ffe7926e512cde1938846af283ae7f199446e4b36bc88892cc1cb","abstract_canon_sha256":"30b0a154acb06d63ce020614444c91142704b7bc49bd8cac6c7a4b98f9c44cae"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-02T01:03:44.359939Z","signature_b64":"c+B3eMQ5SmsZuyQPrPJkNSZ5ftzxRUCPrim89M+8kVOfxA6WjGZBZ8jb5IURRi+SWJLhWCndgSYZfW4dk1npDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"810240108edc486abd013e1eb2bb48014571dbdd3e6a1b128b770e1ee635a1a9","last_reissued_at":"2026-06-02T01:03:44.359417Z","signature_status":"signed_v1","first_computed_at":"2026-06-02T01:03:44.359417Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Constitutional Black-Box Monitoring for Scheming in LLM Agents","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"James Peters-Gill, Marius Hobbhahn, Rich Barton-Cooper, Simon Storf","submitted_at":"2026-02-28T22:31:32Z","abstract_excerpt":"Safe deployment of Large Language Model (LLM) agents in autonomous settings requires reliable oversight mechanisms. A central challenge is detecting scheming, where agents covertly pursue misaligned goals. One approach to mitigating such risks is LLM-based monitoring: using language models to examine agent behaviors for suspicious actions. We study constitutional black-box monitors: prompted classifiers that detect scheming using only externally observable inputs and outputs, optimized on synthetic data generated from natural-language behavior specifications. We introduce two pipelines for gen"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2603.00829","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/2603.00829/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":"2603.00829","created_at":"2026-06-02T01:03:44.359481+00:00"},{"alias_kind":"arxiv_version","alias_value":"2603.00829v2","created_at":"2026-06-02T01:03:44.359481+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2603.00829","created_at":"2026-06-02T01:03:44.359481+00:00"},{"alias_kind":"pith_short_12","alias_value":"QEBEAEEO3REG","created_at":"2026-06-02T01:03:44.359481+00:00"},{"alias_kind":"pith_short_16","alias_value":"QEBEAEEO3REGVPIB","created_at":"2026-06-02T01:03:44.359481+00:00"},{"alias_kind":"pith_short_8","alias_value":"QEBEAEEO","created_at":"2026-06-02T01:03:44.359481+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2605.19270","citing_title":"DECOR: Auditing LLM Deception via Information Manipulation Theory","ref_index":17,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/QEBEAEEO3REGVPIBHYPLFO2IAF","json":"https://pith.science/pith/QEBEAEEO3REGVPIBHYPLFO2IAF.json","graph_json":"https://pith.science/api/pith-number/QEBEAEEO3REGVPIBHYPLFO2IAF/graph.json","events_json":"https://pith.science/api/pith-number/QEBEAEEO3REGVPIBHYPLFO2IAF/events.json","paper":"https://pith.science/paper/QEBEAEEO"},"agent_actions":{"view_html":"https://pith.science/pith/QEBEAEEO3REGVPIBHYPLFO2IAF","download_json":"https://pith.science/pith/QEBEAEEO3REGVPIBHYPLFO2IAF.json","view_paper":"https://pith.science/paper/QEBEAEEO","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2603.00829&json=true","fetch_graph":"https://pith.science/api/pith-number/QEBEAEEO3REGVPIBHYPLFO2IAF/graph.json","fetch_events":"https://pith.science/api/pith-number/QEBEAEEO3REGVPIBHYPLFO2IAF/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/QEBEAEEO3REGVPIBHYPLFO2IAF/action/timestamp_anchor","attest_storage":"https://pith.science/pith/QEBEAEEO3REGVPIBHYPLFO2IAF/action/storage_attestation","attest_author":"https://pith.science/pith/QEBEAEEO3REGVPIBHYPLFO2IAF/action/author_attestation","sign_citation":"https://pith.science/pith/QEBEAEEO3REGVPIBHYPLFO2IAF/action/citation_signature","submit_replication":"https://pith.science/pith/QEBEAEEO3REGVPIBHYPLFO2IAF/action/replication_record"}},"created_at":"2026-06-02T01:03:44.359481+00:00","updated_at":"2026-06-02T01:03:44.359481+00:00"}