{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2024:N3JMGEH4EFUOR5BL6JOGKBB37B","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":"fe9356069bd1329683f545298770f6b48deea3964760833de689a0065543951e","cross_cats_sorted":["cs.CR"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2024-12-20T20:23:52Z","title_canon_sha256":"73d1ab81143f56789d60de6e9940fff2502261f52fa39cc3e878fb5d6b19e789"},"schema_version":"1.0","source":{"id":"2412.16325","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2412.16325","created_at":"2026-07-05T09:52:44Z"},{"alias_kind":"arxiv_version","alias_value":"2412.16325v1","created_at":"2026-07-05T09:52:44Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2412.16325","created_at":"2026-07-05T09:52:44Z"},{"alias_kind":"pith_short_12","alias_value":"N3JMGEH4EFUO","created_at":"2026-07-05T09:52:44Z"},{"alias_kind":"pith_short_16","alias_value":"N3JMGEH4EFUOR5BL","created_at":"2026-07-05T09:52:44Z"},{"alias_kind":"pith_short_8","alias_value":"N3JMGEH4","created_at":"2026-07-05T09:52:44Z"}],"graph_snapshots":[{"event_id":"sha256:f6bed8e84eb8cf6516bb64dbaacb86b4c02a5335aa793f6ced99de28a4523aac","target":"graph","created_at":"2026-07-05T09:52:44Z","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/2412.16325/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"As AI systems increasingly make critical decisions, deceptive AI poses a significant challenge to trust and safety. We present Self-Other Overlap (SOO) fine-tuning, a promising approach in AI Safety that could substantially improve our ability to build honest artificial intelligence. Inspired by cognitive neuroscience research on empathy, SOO aims to align how AI models represent themselves and others. Our experiments on LLMs with 7B, 27B, and 78B parameters demonstrate SOO's efficacy: deceptive responses of Mistral-7B-Instruct-v0.2 dropped from 73.6% to 17.2% with no observed reduction in gen","authors_text":"Cameron Berg, Diogo Schwerz de Lucena, Judd Rosenblatt, Marc Carauleanu, Michael Vaiana","cross_cats":["cs.CR"],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2024-12-20T20:23:52Z","title":"Towards Safe and Honest AI Agents with Neural Self-Other Overlap"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2412.16325","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:b8259a9ae3356214c3ed692b9bdf4735242ee5b69637cc0d8260a503b784d48d","target":"record","created_at":"2026-07-05T09:52:44Z","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":"fe9356069bd1329683f545298770f6b48deea3964760833de689a0065543951e","cross_cats_sorted":["cs.CR"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2024-12-20T20:23:52Z","title_canon_sha256":"73d1ab81143f56789d60de6e9940fff2502261f52fa39cc3e878fb5d6b19e789"},"schema_version":"1.0","source":{"id":"2412.16325","kind":"arxiv","version":1}},"canonical_sha256":"6ed2c310fc2168e8f42bf25c65043bf84836c61719a799ee1bbdca8b8359f7d2","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"6ed2c310fc2168e8f42bf25c65043bf84836c61719a799ee1bbdca8b8359f7d2","first_computed_at":"2026-07-05T09:52:44.642466Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T09:52:44.642466Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"Ry+GzTv7epRD2tG0Y1SCSuGUU7ojlclGI5aYb3+RklqSR8Me/uTd4M2CQ7Ae7fuc6CcdsRin2nBq2xiBnEi0Ag==","signature_status":"signed_v1","signed_at":"2026-07-05T09:52:44.642961Z","signed_message":"canonical_sha256_bytes"},"source_id":"2412.16325","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:b8259a9ae3356214c3ed692b9bdf4735242ee5b69637cc0d8260a503b784d48d","sha256:f6bed8e84eb8cf6516bb64dbaacb86b4c02a5335aa793f6ced99de28a4523aac"],"state_sha256":"954f533b14d4d21e097a967e68e8c887901ac195802e7c008dc730ef480e02ff"}