{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:7K5SDFW4TEWNMR3RESA3GNLNUZ","short_pith_number":"pith:7K5SDFW4","schema_version":"1.0","canonical_sha256":"fabb2196dc992cd647712481b3356da66a5c148cc26902dd100ebb20c07b45f3","source":{"kind":"arxiv","id":"2605.29601","version":1},"attestation_state":"computed","paper":{"title":"Training Deliberative Monitors for Black-Box Scheming Detection","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Aditya Sinha, Akshat Naik, Axel H{\\o}jmark, Kilian Merkelbach, Marius Hobbhahn, Rich Barton-Cooper, Simon Storf, Victor Gillioz","submitted_at":"2026-05-28T08:40:32Z","abstract_excerpt":"As autonomous agents become more capable of performing real-world tasks, distinguishing scheming behavior from benign task pursuit may become a central AI control problem. Existing monitors often rely on chain-of-thought access or internal activations, or use prompted frontier models, all of which can be unavailable, unreliable or expensive in deployment. In this work, we study action-only deliberative monitors: smaller open-weight models trained to detect scheming and sabotage from agentic trajectories without accessing the monitored agent's reasoning or model internals. Our method, inspired "},"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":"2605.29601","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-05-28T08:40:32Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"fcd2bf5845c63e6a81744dc67fc39897865ad456e6ae634e985104821242a6d2","abstract_canon_sha256":"78b8dd90a764453c74b03f260af7f204dc7fd4047473671f3d7a999436a1219c"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-29T01:05:50.343526Z","signature_b64":"fS+9dT6Fxf8RjxU2R21KoplSIeqP4ZwxBTkUVpYO5JT1ktEZZItr3eqo6SM/gS37LVcdjw2PgEjUgX1Tb+w7Ag==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"fabb2196dc992cd647712481b3356da66a5c148cc26902dd100ebb20c07b45f3","last_reissued_at":"2026-05-29T01:05:50.342923Z","signature_status":"signed_v1","first_computed_at":"2026-05-29T01:05:50.342923Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Training Deliberative Monitors for Black-Box Scheming Detection","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Aditya Sinha, Akshat Naik, Axel H{\\o}jmark, Kilian Merkelbach, Marius Hobbhahn, Rich Barton-Cooper, Simon Storf, Victor Gillioz","submitted_at":"2026-05-28T08:40:32Z","abstract_excerpt":"As autonomous agents become more capable of performing real-world tasks, distinguishing scheming behavior from benign task pursuit may become a central AI control problem. Existing monitors often rely on chain-of-thought access or internal activations, or use prompted frontier models, all of which can be unavailable, unreliable or expensive in deployment. In this work, we study action-only deliberative monitors: smaller open-weight models trained to detect scheming and sabotage from agentic trajectories without accessing the monitored agent's reasoning or model internals. Our method, inspired "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.29601","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/2605.29601/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":"2605.29601","created_at":"2026-05-29T01:05:50.343006+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.29601v1","created_at":"2026-05-29T01:05:50.343006+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.29601","created_at":"2026-05-29T01:05:50.343006+00:00"},{"alias_kind":"pith_short_12","alias_value":"7K5SDFW4TEWN","created_at":"2026-05-29T01:05:50.343006+00:00"},{"alias_kind":"pith_short_16","alias_value":"7K5SDFW4TEWNMR3R","created_at":"2026-05-29T01:05:50.343006+00:00"},{"alias_kind":"pith_short_8","alias_value":"7K5SDFW4","created_at":"2026-05-29T01:05:50.343006+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/7K5SDFW4TEWNMR3RESA3GNLNUZ","json":"https://pith.science/pith/7K5SDFW4TEWNMR3RESA3GNLNUZ.json","graph_json":"https://pith.science/api/pith-number/7K5SDFW4TEWNMR3RESA3GNLNUZ/graph.json","events_json":"https://pith.science/api/pith-number/7K5SDFW4TEWNMR3RESA3GNLNUZ/events.json","paper":"https://pith.science/paper/7K5SDFW4"},"agent_actions":{"view_html":"https://pith.science/pith/7K5SDFW4TEWNMR3RESA3GNLNUZ","download_json":"https://pith.science/pith/7K5SDFW4TEWNMR3RESA3GNLNUZ.json","view_paper":"https://pith.science/paper/7K5SDFW4","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.29601&json=true","fetch_graph":"https://pith.science/api/pith-number/7K5SDFW4TEWNMR3RESA3GNLNUZ/graph.json","fetch_events":"https://pith.science/api/pith-number/7K5SDFW4TEWNMR3RESA3GNLNUZ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/7K5SDFW4TEWNMR3RESA3GNLNUZ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/7K5SDFW4TEWNMR3RESA3GNLNUZ/action/storage_attestation","attest_author":"https://pith.science/pith/7K5SDFW4TEWNMR3RESA3GNLNUZ/action/author_attestation","sign_citation":"https://pith.science/pith/7K5SDFW4TEWNMR3RESA3GNLNUZ/action/citation_signature","submit_replication":"https://pith.science/pith/7K5SDFW4TEWNMR3RESA3GNLNUZ/action/replication_record"}},"created_at":"2026-05-29T01:05:50.343006+00:00","updated_at":"2026-05-29T01:05:50.343006+00:00"}