{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:EESR5EGJT2MVGAXDK6NRJSVSNP","short_pith_number":"pith:EESR5EGJ","schema_version":"1.0","canonical_sha256":"21251e90c99e995302e3579b14cab26be24e94dc606ccf5c32dba664a32bac68","source":{"kind":"arxiv","id":"2607.01567","version":1},"attestation_state":"computed","paper":{"title":"Scaling Trends for Lie Detector Oversight in Preference Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Adam Gleave, Ann-Kathrin Dombrowski, Chris Cundy, Oskar J. Hollinsworth, Sam Adam-Day","submitted_at":"2026-07-02T00:53:33Z","abstract_excerpt":"Deceptive behavior in LLMs is costly to monitor and prevent, motivating approaches such as Scalable Oversight via Lie Detectors (SOLiD) (Cundy & Gleave, 2025), which uses lie detectors to identify responses for review by high-cost labelers. In this paper, we scale SOLiD to larger models and evaluate it in more diverse and realistic preference-learning settings. We find favorable scaling: undetected deception drops from 34% for 1B-parameter models to 14% for 405B-parameter models at a detector true positive rate of 99%, and expensive human labelers can be removed entirely from the fine-tuning p"},"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":"2607.01567","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2026-07-02T00:53:33Z","cross_cats_sorted":[],"title_canon_sha256":"a85355b3641376327c7789ba918f352c07732f9dd62208d1d1255221760c6b22","abstract_canon_sha256":"b8dce63bba6417fbc4217141b381ea10c40c9b957b34888495cfba0628b7cc01"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-03T01:17:02.072721Z","signature_b64":"VqIPIe8lRftn603X313VsP5lG92f8cSMFKQrIgv1V5N4/EReNulGvo9tG2aLV33GVsO+1eE2WkmfhPohVFBDAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"21251e90c99e995302e3579b14cab26be24e94dc606ccf5c32dba664a32bac68","last_reissued_at":"2026-07-03T01:17:02.072209Z","signature_status":"signed_v1","first_computed_at":"2026-07-03T01:17:02.072209Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Scaling Trends for Lie Detector Oversight in Preference Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Adam Gleave, Ann-Kathrin Dombrowski, Chris Cundy, Oskar J. Hollinsworth, Sam Adam-Day","submitted_at":"2026-07-02T00:53:33Z","abstract_excerpt":"Deceptive behavior in LLMs is costly to monitor and prevent, motivating approaches such as Scalable Oversight via Lie Detectors (SOLiD) (Cundy & Gleave, 2025), which uses lie detectors to identify responses for review by high-cost labelers. In this paper, we scale SOLiD to larger models and evaluate it in more diverse and realistic preference-learning settings. We find favorable scaling: undetected deception drops from 34% for 1B-parameter models to 14% for 405B-parameter models at a detector true positive rate of 99%, and expensive human labelers can be removed entirely from the fine-tuning p"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2607.01567","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/2607.01567/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":"2607.01567","created_at":"2026-07-03T01:17:02.072271+00:00"},{"alias_kind":"arxiv_version","alias_value":"2607.01567v1","created_at":"2026-07-03T01:17:02.072271+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2607.01567","created_at":"2026-07-03T01:17:02.072271+00:00"},{"alias_kind":"pith_short_12","alias_value":"EESR5EGJT2MV","created_at":"2026-07-03T01:17:02.072271+00:00"},{"alias_kind":"pith_short_16","alias_value":"EESR5EGJT2MVGAXD","created_at":"2026-07-03T01:17:02.072271+00:00"},{"alias_kind":"pith_short_8","alias_value":"EESR5EGJ","created_at":"2026-07-03T01:17:02.072271+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/EESR5EGJT2MVGAXDK6NRJSVSNP","json":"https://pith.science/pith/EESR5EGJT2MVGAXDK6NRJSVSNP.json","graph_json":"https://pith.science/api/pith-number/EESR5EGJT2MVGAXDK6NRJSVSNP/graph.json","events_json":"https://pith.science/api/pith-number/EESR5EGJT2MVGAXDK6NRJSVSNP/events.json","paper":"https://pith.science/paper/EESR5EGJ"},"agent_actions":{"view_html":"https://pith.science/pith/EESR5EGJT2MVGAXDK6NRJSVSNP","download_json":"https://pith.science/pith/EESR5EGJT2MVGAXDK6NRJSVSNP.json","view_paper":"https://pith.science/paper/EESR5EGJ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2607.01567&json=true","fetch_graph":"https://pith.science/api/pith-number/EESR5EGJT2MVGAXDK6NRJSVSNP/graph.json","fetch_events":"https://pith.science/api/pith-number/EESR5EGJT2MVGAXDK6NRJSVSNP/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/EESR5EGJT2MVGAXDK6NRJSVSNP/action/timestamp_anchor","attest_storage":"https://pith.science/pith/EESR5EGJT2MVGAXDK6NRJSVSNP/action/storage_attestation","attest_author":"https://pith.science/pith/EESR5EGJT2MVGAXDK6NRJSVSNP/action/author_attestation","sign_citation":"https://pith.science/pith/EESR5EGJT2MVGAXDK6NRJSVSNP/action/citation_signature","submit_replication":"https://pith.science/pith/EESR5EGJT2MVGAXDK6NRJSVSNP/action/replication_record"}},"created_at":"2026-07-03T01:17:02.072271+00:00","updated_at":"2026-07-03T01:17:02.072271+00:00"}