{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:BOXKLDBX3S62RA6JFN2QMDX3RV","short_pith_number":"pith:BOXKLDBX","schema_version":"1.0","canonical_sha256":"0baea58c37dcbda883c92b75060efb8d76cfef599ef3bd7df470a68c763e3f10","source":{"kind":"arxiv","id":"2510.06096","version":3},"attestation_state":"computed","paper":{"title":"The Alignment Auditor: A Bayesian Framework for Verifying and Refining LLM Objectives","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Arjun Jagota, Matthieu Bou, Nyal Patel, Satyapriya Krishna, Sonali Parbhoo","submitted_at":"2025-10-07T16:25:14Z","abstract_excerpt":"The objectives that Large Language Models (LLMs) implicitly optimize remain dangerously opaque, making trustworthy alignment and auditing a grand challenge. While Inverse Reinforcement Learning (IRL) can infer reward functions from behaviour, existing approaches either produce a single, overconfident reward estimate or fail to address the fundamental ambiguity of the task (non-identifiability). This paper introduces a principled auditing framework that re-frames reward inference from a simple estimation task to a comprehensive process for verification. Our framework leverages Bayesian IRL to n"},"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":"2510.06096","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-10-07T16:25:14Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"47b650cf8d683bb7c8eb6763cc36fe77ad0c461429d9fc3f20ce108103c0d213","abstract_canon_sha256":"dcc690cd91f80aa0e60348b3917d130006e5f895db54972f6249ab37a358fe12"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-30T00:15:08.137157Z","signature_b64":"YCLupTpuQkEPBk4AMxPDOhzDnX6gnK24cL6MSvrWHZk9Ad2HhqXs6dtc+zSlVIytnyi5h5eoNA7TnwA0r/3xCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0baea58c37dcbda883c92b75060efb8d76cfef599ef3bd7df470a68c763e3f10","last_reissued_at":"2026-06-30T00:15:08.136671Z","signature_status":"signed_v1","first_computed_at":"2026-06-30T00:15:08.136671Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"The Alignment Auditor: A Bayesian Framework for Verifying and Refining LLM Objectives","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Arjun Jagota, Matthieu Bou, Nyal Patel, Satyapriya Krishna, Sonali Parbhoo","submitted_at":"2025-10-07T16:25:14Z","abstract_excerpt":"The objectives that Large Language Models (LLMs) implicitly optimize remain dangerously opaque, making trustworthy alignment and auditing a grand challenge. While Inverse Reinforcement Learning (IRL) can infer reward functions from behaviour, existing approaches either produce a single, overconfident reward estimate or fail to address the fundamental ambiguity of the task (non-identifiability). This paper introduces a principled auditing framework that re-frames reward inference from a simple estimation task to a comprehensive process for verification. Our framework leverages Bayesian IRL to n"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2510.06096","kind":"arxiv","version":3},"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/2510.06096/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":"2510.06096","created_at":"2026-06-30T00:15:08.136721+00:00"},{"alias_kind":"arxiv_version","alias_value":"2510.06096v3","created_at":"2026-06-30T00:15:08.136721+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2510.06096","created_at":"2026-06-30T00:15:08.136721+00:00"},{"alias_kind":"pith_short_12","alias_value":"BOXKLDBX3S62","created_at":"2026-06-30T00:15:08.136721+00:00"},{"alias_kind":"pith_short_16","alias_value":"BOXKLDBX3S62RA6J","created_at":"2026-06-30T00:15:08.136721+00:00"},{"alias_kind":"pith_short_8","alias_value":"BOXKLDBX","created_at":"2026-06-30T00:15:08.136721+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/BOXKLDBX3S62RA6JFN2QMDX3RV","json":"https://pith.science/pith/BOXKLDBX3S62RA6JFN2QMDX3RV.json","graph_json":"https://pith.science/api/pith-number/BOXKLDBX3S62RA6JFN2QMDX3RV/graph.json","events_json":"https://pith.science/api/pith-number/BOXKLDBX3S62RA6JFN2QMDX3RV/events.json","paper":"https://pith.science/paper/BOXKLDBX"},"agent_actions":{"view_html":"https://pith.science/pith/BOXKLDBX3S62RA6JFN2QMDX3RV","download_json":"https://pith.science/pith/BOXKLDBX3S62RA6JFN2QMDX3RV.json","view_paper":"https://pith.science/paper/BOXKLDBX","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2510.06096&json=true","fetch_graph":"https://pith.science/api/pith-number/BOXKLDBX3S62RA6JFN2QMDX3RV/graph.json","fetch_events":"https://pith.science/api/pith-number/BOXKLDBX3S62RA6JFN2QMDX3RV/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/BOXKLDBX3S62RA6JFN2QMDX3RV/action/timestamp_anchor","attest_storage":"https://pith.science/pith/BOXKLDBX3S62RA6JFN2QMDX3RV/action/storage_attestation","attest_author":"https://pith.science/pith/BOXKLDBX3S62RA6JFN2QMDX3RV/action/author_attestation","sign_citation":"https://pith.science/pith/BOXKLDBX3S62RA6JFN2QMDX3RV/action/citation_signature","submit_replication":"https://pith.science/pith/BOXKLDBX3S62RA6JFN2QMDX3RV/action/replication_record"}},"created_at":"2026-06-30T00:15:08.136721+00:00","updated_at":"2026-06-30T00:15:08.136721+00:00"}