{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2025:YDGB4BF2C2TASQ3TV7ZGTI6XIN","short_pith_number":"pith:YDGB4BF2","canonical_record":{"source":{"id":"2502.03407","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2025-02-05T17:49:40Z","cross_cats_sorted":[],"title_canon_sha256":"35c407a799af75675b9ade8c66efb76ccd4c1fe9c711485f2b734add52899a40","abstract_canon_sha256":"596bb947e33c992c61eb269cc7ed9ae681202a623f90ff17f7917dba57b828e8"},"schema_version":"1.0"},"canonical_sha256":"c0cc1e04ba16a6094373aff269a3d7436d4945a4ca4ee11e0828f91b24f7ab06","source":{"kind":"arxiv","id":"2502.03407","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2502.03407","created_at":"2026-07-05T10:10:04Z"},{"alias_kind":"arxiv_version","alias_value":"2502.03407v1","created_at":"2026-07-05T10:10:04Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.03407","created_at":"2026-07-05T10:10:04Z"},{"alias_kind":"pith_short_12","alias_value":"YDGB4BF2C2TA","created_at":"2026-07-05T10:10:04Z"},{"alias_kind":"pith_short_16","alias_value":"YDGB4BF2C2TASQ3T","created_at":"2026-07-05T10:10:04Z"},{"alias_kind":"pith_short_8","alias_value":"YDGB4BF2","created_at":"2026-07-05T10:10:04Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2025:YDGB4BF2C2TASQ3TV7ZGTI6XIN","target":"record","payload":{"canonical_record":{"source":{"id":"2502.03407","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2025-02-05T17:49:40Z","cross_cats_sorted":[],"title_canon_sha256":"35c407a799af75675b9ade8c66efb76ccd4c1fe9c711485f2b734add52899a40","abstract_canon_sha256":"596bb947e33c992c61eb269cc7ed9ae681202a623f90ff17f7917dba57b828e8"},"schema_version":"1.0"},"canonical_sha256":"c0cc1e04ba16a6094373aff269a3d7436d4945a4ca4ee11e0828f91b24f7ab06","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:10:04.473142Z","signature_b64":"nQRyIKC4TZmfIU1gnKryrso40JiWNmLUAsw5jSvs1JWIs3r8CcEC8UmUU1UhUTLWN+AuiRfHjbhozFF6psyNAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"c0cc1e04ba16a6094373aff269a3d7436d4945a4ca4ee11e0828f91b24f7ab06","last_reissued_at":"2026-07-05T10:10:04.472641Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:10:04.472641Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2502.03407","source_version":1,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T10:10:04Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"Ux6JJCOk3aROTxzIzquNJQKtaL5s9slcH8LDjERk5hnkAzJ0LyxSkfY6C4+vSH/MUvsl3erqMXjxMg1BipG2CQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-06T06:24:23.630696Z"},"content_sha256":"b219c7fb95adc011c188b81fb9882df38963b4b1857c10abe12d7b01547f43eb","schema_version":"1.0","event_id":"sha256:b219c7fb95adc011c188b81fb9882df38963b4b1857c10abe12d7b01547f43eb"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2025:YDGB4BF2C2TASQ3TV7ZGTI6XIN","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Detecting Strategic Deception Using Linear Probes","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Bilal Chughtai, Marius Hobbhahn, Nicholas Goldowsky-Dill, Stefan Heimersheim","submitted_at":"2025-02-05T17:49:40Z","abstract_excerpt":"AI models might use deceptive strategies as part of scheming or misaligned behaviour. Monitoring outputs alone is insufficient, since the AI might produce seemingly benign outputs while their internal reasoning is misaligned. We thus evaluate if linear probes can robustly detect deception by monitoring model activations. We test two probe-training datasets, one with contrasting instructions to be honest or deceptive (following Zou et al., 2023) and one of responses to simple roleplaying scenarios. We test whether these probes generalize to realistic settings where Llama-3.3-70B-Instruct behave"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.03407","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/2502.03407/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"},"verdict_id":null},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T10:10:04Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"QGdunX5KKNXiYVUDexqcmX5pYKOui5bhyOQFYB8RibZbh7TpvttucwEdYtPksfnuiVsa/dZ7FSaiDzbJ49G4Bg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-06T06:24:23.631206Z"},"content_sha256":"c7754cda91819bcdcd463edc454df0f785fd84c97d374637cdf6145500468124","schema_version":"1.0","event_id":"sha256:c7754cda91819bcdcd463edc454df0f785fd84c97d374637cdf6145500468124"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/YDGB4BF2C2TASQ3TV7ZGTI6XIN/bundle.json","state_url":"https://pith.science/pith/YDGB4BF2C2TASQ3TV7ZGTI6XIN/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/YDGB4BF2C2TASQ3TV7ZGTI6XIN/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-08-06T06:24:23Z","links":{"resolver":"https://pith.science/pith/YDGB4BF2C2TASQ3TV7ZGTI6XIN","bundle":"https://pith.science/pith/YDGB4BF2C2TASQ3TV7ZGTI6XIN/bundle.json","state":"https://pith.science/pith/YDGB4BF2C2TASQ3TV7ZGTI6XIN/state.json","well_known_bundle":"https://pith.science/.well-known/pith/YDGB4BF2C2TASQ3TV7ZGTI6XIN/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2025:YDGB4BF2C2TASQ3TV7ZGTI6XIN","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"596bb947e33c992c61eb269cc7ed9ae681202a623f90ff17f7917dba57b828e8","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2025-02-05T17:49:40Z","title_canon_sha256":"35c407a799af75675b9ade8c66efb76ccd4c1fe9c711485f2b734add52899a40"},"schema_version":"1.0","source":{"id":"2502.03407","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2502.03407","created_at":"2026-07-05T10:10:04Z"},{"alias_kind":"arxiv_version","alias_value":"2502.03407v1","created_at":"2026-07-05T10:10:04Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.03407","created_at":"2026-07-05T10:10:04Z"},{"alias_kind":"pith_short_12","alias_value":"YDGB4BF2C2TA","created_at":"2026-07-05T10:10:04Z"},{"alias_kind":"pith_short_16","alias_value":"YDGB4BF2C2TASQ3T","created_at":"2026-07-05T10:10:04Z"},{"alias_kind":"pith_short_8","alias_value":"YDGB4BF2","created_at":"2026-07-05T10:10:04Z"}],"graph_snapshots":[{"event_id":"sha256:c7754cda91819bcdcd463edc454df0f785fd84c97d374637cdf6145500468124","target":"graph","created_at":"2026-07-05T10:10:04Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2502.03407/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"AI models might use deceptive strategies as part of scheming or misaligned behaviour. Monitoring outputs alone is insufficient, since the AI might produce seemingly benign outputs while their internal reasoning is misaligned. We thus evaluate if linear probes can robustly detect deception by monitoring model activations. We test two probe-training datasets, one with contrasting instructions to be honest or deceptive (following Zou et al., 2023) and one of responses to simple roleplaying scenarios. We test whether these probes generalize to realistic settings where Llama-3.3-70B-Instruct behave","authors_text":"Bilal Chughtai, Marius Hobbhahn, Nicholas Goldowsky-Dill, Stefan Heimersheim","cross_cats":[],"headline":"","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2025-02-05T17:49:40Z","title":"Detecting Strategic Deception Using Linear Probes"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.03407","kind":"arxiv","version":1},"verdict":{"created_at":null,"id":null,"model_set":{},"one_line_summary":"","pipeline_version":null,"pith_extraction_headline":"","strongest_claim":"","weakest_assumption":""}},"verdict_id":null}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:b219c7fb95adc011c188b81fb9882df38963b4b1857c10abe12d7b01547f43eb","target":"record","created_at":"2026-07-05T10:10:04Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"596bb947e33c992c61eb269cc7ed9ae681202a623f90ff17f7917dba57b828e8","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2025-02-05T17:49:40Z","title_canon_sha256":"35c407a799af75675b9ade8c66efb76ccd4c1fe9c711485f2b734add52899a40"},"schema_version":"1.0","source":{"id":"2502.03407","kind":"arxiv","version":1}},"canonical_sha256":"c0cc1e04ba16a6094373aff269a3d7436d4945a4ca4ee11e0828f91b24f7ab06","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"c0cc1e04ba16a6094373aff269a3d7436d4945a4ca4ee11e0828f91b24f7ab06","first_computed_at":"2026-07-05T10:10:04.472641Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T10:10:04.472641Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"nQRyIKC4TZmfIU1gnKryrso40JiWNmLUAsw5jSvs1JWIs3r8CcEC8UmUU1UhUTLWN+AuiRfHjbhozFF6psyNAQ==","signature_status":"signed_v1","signed_at":"2026-07-05T10:10:04.473142Z","signed_message":"canonical_sha256_bytes"},"source_id":"2502.03407","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:b219c7fb95adc011c188b81fb9882df38963b4b1857c10abe12d7b01547f43eb","sha256:c7754cda91819bcdcd463edc454df0f785fd84c97d374637cdf6145500468124"],"state_sha256":"91072c0545ddf0a87d04fd0960942ac8d60af878d02852763da8a802fd75af15"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"SqINe6z/E85kR9St//IoH3qBvSDl10kNjL2oRAEYOwijStbciSxhi7lnmCAepx8beQoKQXWhEEbC23U5yvXtDQ==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-06T06:24:23.637775Z","bundle_sha256":"190379fabbc26b93cdb02210f83b0da156a331b13beb6084caa6249b0f8f8447"}}