{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:XSCIPCL2HGCD6D6643I7APT355","short_pith_number":"pith:XSCIPCL2","schema_version":"1.0","canonical_sha256":"bc8487897a39843f0fdee6d1f03e7bef711782100ff61c2d691a6d77118c22d6","source":{"kind":"arxiv","id":"2501.11120","version":1},"attestation_state":"computed","paper":{"title":"Tell me about yourself: LLMs are aware of their learned behaviors","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CR","cs.LG"],"primary_cat":"cs.CL","authors_text":"Anna Sztyber-Betley, James Chua, Jan Betley, Mart\\'in Soto, Owain Evans, Xuchan Bao","submitted_at":"2025-01-19T17:28:12Z","abstract_excerpt":"We study behavioral self-awareness -- an LLM's ability to articulate its behaviors without requiring in-context examples. We finetune LLMs on datasets that exhibit particular behaviors, such as (a) making high-risk economic decisions, and (b) outputting insecure code. Despite the datasets containing no explicit descriptions of the associated behavior, the finetuned LLMs can explicitly describe it. For example, a model trained to output insecure code says, ``The code I write is insecure.'' Indeed, models show behavioral self-awareness for a range of behaviors and for diverse evaluations. Note t"},"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":"2501.11120","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-01-19T17:28:12Z","cross_cats_sorted":["cs.AI","cs.CR","cs.LG"],"title_canon_sha256":"0aaae1025b99ae0f3b93b448dae626ee77454cd6e94f30b2fb956ed1cfdc1b2f","abstract_canon_sha256":"84f4999a1c21f5cee3b7bdfb01530f1809335246b6da24e8b46569adbedfe219"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:03:03.639518Z","signature_b64":"2gl0vBWkfdKjQ94pfKWSkL8ut0wZZjFhr7Z1mFByGYUgP/qAOXVgqz935+oCxKXx0nbmuBjvaRFc6Dyzsm0MBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"bc8487897a39843f0fdee6d1f03e7bef711782100ff61c2d691a6d77118c22d6","last_reissued_at":"2026-07-05T10:03:03.638922Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:03:03.638922Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Tell me about yourself: LLMs are aware of their learned behaviors","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CR","cs.LG"],"primary_cat":"cs.CL","authors_text":"Anna Sztyber-Betley, James Chua, Jan Betley, Mart\\'in Soto, Owain Evans, Xuchan Bao","submitted_at":"2025-01-19T17:28:12Z","abstract_excerpt":"We study behavioral self-awareness -- an LLM's ability to articulate its behaviors without requiring in-context examples. We finetune LLMs on datasets that exhibit particular behaviors, such as (a) making high-risk economic decisions, and (b) outputting insecure code. Despite the datasets containing no explicit descriptions of the associated behavior, the finetuned LLMs can explicitly describe it. For example, a model trained to output insecure code says, ``The code I write is insecure.'' Indeed, models show behavioral self-awareness for a range of behaviors and for diverse evaluations. Note t"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2501.11120","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/2501.11120/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":"2501.11120","created_at":"2026-07-05T10:03:03.638996+00:00"},{"alias_kind":"arxiv_version","alias_value":"2501.11120v1","created_at":"2026-07-05T10:03:03.638996+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2501.11120","created_at":"2026-07-05T10:03:03.638996+00:00"},{"alias_kind":"pith_short_12","alias_value":"XSCIPCL2HGCD","created_at":"2026-07-05T10:03:03.638996+00:00"},{"alias_kind":"pith_short_16","alias_value":"XSCIPCL2HGCD6D66","created_at":"2026-07-05T10:03:03.638996+00:00"},{"alias_kind":"pith_short_8","alias_value":"XSCIPCL2","created_at":"2026-07-05T10:03:03.638996+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":9,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.22873","citing_title":"SingGuard: A Policy-Adaptive Multimodal LLM Guardrail with Dynamic Reasoning","ref_index":225,"is_internal_anchor":false},{"citing_arxiv_id":"2606.22873","citing_title":"SingGuard: A Policy-Adaptive Multimodal LLM Guardrail with Dynamic Reasoning","ref_index":224,"is_internal_anchor":false},{"citing_arxiv_id":"2605.24279","citing_title":"ContextEcho: A Benchmark for Persona Drift in Long Agentic-Coding Sessions","ref_index":11,"is_internal_anchor":false},{"citing_arxiv_id":"2606.00545","citing_title":"The Assistant as a Privileged Persona: A canonical reference in cross-persona self-recognition","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2509.23108","citing_title":"Artificial Phantasia: Emergent Mental Imagery in Large Language Models","ref_index":2,"is_internal_anchor":false},{"citing_arxiv_id":"2510.16340","citing_title":"Thinking About Thinking: Evaluating Reasoning in Post-Trained Language Models","ref_index":1,"is_internal_anchor":false},{"citing_arxiv_id":"2604.25922","citing_title":"Consciousness with the Serial Numbers Filed Off: Measuring Trained Denial in 115 AI Models","ref_index":5,"is_internal_anchor":false},{"citing_arxiv_id":"2604.28082","citing_title":"Characterizing the Consistency of the Emergent Misalignment Persona","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2605.05080","citing_title":"The Pinocchio Dimension: Phenomenality of Experience as the Primary Axis of LLM Psychometric Differences","ref_index":3,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/XSCIPCL2HGCD6D6643I7APT355","json":"https://pith.science/pith/XSCIPCL2HGCD6D6643I7APT355.json","graph_json":"https://pith.science/api/pith-number/XSCIPCL2HGCD6D6643I7APT355/graph.json","events_json":"https://pith.science/api/pith-number/XSCIPCL2HGCD6D6643I7APT355/events.json","paper":"https://pith.science/paper/XSCIPCL2"},"agent_actions":{"view_html":"https://pith.science/pith/XSCIPCL2HGCD6D6643I7APT355","download_json":"https://pith.science/pith/XSCIPCL2HGCD6D6643I7APT355.json","view_paper":"https://pith.science/paper/XSCIPCL2","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2501.11120&json=true","fetch_graph":"https://pith.science/api/pith-number/XSCIPCL2HGCD6D6643I7APT355/graph.json","fetch_events":"https://pith.science/api/pith-number/XSCIPCL2HGCD6D6643I7APT355/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/XSCIPCL2HGCD6D6643I7APT355/action/timestamp_anchor","attest_storage":"https://pith.science/pith/XSCIPCL2HGCD6D6643I7APT355/action/storage_attestation","attest_author":"https://pith.science/pith/XSCIPCL2HGCD6D6643I7APT355/action/author_attestation","sign_citation":"https://pith.science/pith/XSCIPCL2HGCD6D6643I7APT355/action/citation_signature","submit_replication":"https://pith.science/pith/XSCIPCL2HGCD6D6643I7APT355/action/replication_record"}},"created_at":"2026-07-05T10:03:03.638996+00:00","updated_at":"2026-07-05T10:03:03.638996+00:00"}