{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:UDVBFP75RR6HELUNCKOTVXNLYJ","short_pith_number":"pith:UDVBFP75","schema_version":"1.0","canonical_sha256":"a0ea12bffd8c7c722e8d129d3addabc2784ad3580de8a3d31642a580f0f4bc1c","source":{"kind":"arxiv","id":"2602.08819","version":2},"attestation_state":"computed","paper":{"title":"Bayesian Preference Learning for Test-Time Steerable Reward Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Jiwoo Hong, Shao Tang, Zhipeng Wang","submitted_at":"2026-02-09T15:55:56Z","abstract_excerpt":"Reward models are central to aligning language models with human preferences via reinforcement learning (RL). As RL is increasingly applied to settings such as verifiable rewards and multi-objective alignment, RMs are expected to encode more complex and multifaceted preference distributions. However, classifier RMs remain static once trained, limiting their adaptability at test time. We propose Variational In-Context Reward Modeling (ICRM), a novel Bayesian reward modeling objective that enables test-time steerability via in-context preference demonstrations. ICRM casts reward modeling as amor"},"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":"2602.08819","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-02-09T15:55:56Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"2cc4cf1a6728151fb968eb5520c0eccd4be1f8caf36bdab9af4e3381003bc2b1","abstract_canon_sha256":"e2e963c9b1908d910fae6e3196a2ca5a0e3d3e972151f11fd9d48a9d6463cc78"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-21T01:04:22.653933Z","signature_b64":"Hc2SgI07dzRvD4rGWUMwygao4OtCsv+0aqf6KR11K+X+7cnEExxzprktbzd2BwlILaSy8YCEwKXph2JLm4taCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a0ea12bffd8c7c722e8d129d3addabc2784ad3580de8a3d31642a580f0f4bc1c","last_reissued_at":"2026-05-21T01:04:22.649496Z","signature_status":"signed_v1","first_computed_at":"2026-05-21T01:04:22.649496Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Bayesian Preference Learning for Test-Time Steerable Reward Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Jiwoo Hong, Shao Tang, Zhipeng Wang","submitted_at":"2026-02-09T15:55:56Z","abstract_excerpt":"Reward models are central to aligning language models with human preferences via reinforcement learning (RL). As RL is increasingly applied to settings such as verifiable rewards and multi-objective alignment, RMs are expected to encode more complex and multifaceted preference distributions. However, classifier RMs remain static once trained, limiting their adaptability at test time. We propose Variational In-Context Reward Modeling (ICRM), a novel Bayesian reward modeling objective that enables test-time steerability via in-context preference demonstrations. ICRM casts reward modeling as amor"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2602.08819","kind":"arxiv","version":2},"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/2602.08819/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":"2602.08819","created_at":"2026-05-21T01:04:22.649641+00:00"},{"alias_kind":"arxiv_version","alias_value":"2602.08819v2","created_at":"2026-05-21T01:04:22.649641+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2602.08819","created_at":"2026-05-21T01:04:22.649641+00:00"},{"alias_kind":"pith_short_12","alias_value":"UDVBFP75RR6H","created_at":"2026-05-21T01:04:22.649641+00:00"},{"alias_kind":"pith_short_16","alias_value":"UDVBFP75RR6HELUN","created_at":"2026-05-21T01:04:22.649641+00:00"},{"alias_kind":"pith_short_8","alias_value":"UDVBFP75","created_at":"2026-05-21T01:04:22.649641+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/UDVBFP75RR6HELUNCKOTVXNLYJ","json":"https://pith.science/pith/UDVBFP75RR6HELUNCKOTVXNLYJ.json","graph_json":"https://pith.science/api/pith-number/UDVBFP75RR6HELUNCKOTVXNLYJ/graph.json","events_json":"https://pith.science/api/pith-number/UDVBFP75RR6HELUNCKOTVXNLYJ/events.json","paper":"https://pith.science/paper/UDVBFP75"},"agent_actions":{"view_html":"https://pith.science/pith/UDVBFP75RR6HELUNCKOTVXNLYJ","download_json":"https://pith.science/pith/UDVBFP75RR6HELUNCKOTVXNLYJ.json","view_paper":"https://pith.science/paper/UDVBFP75","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2602.08819&json=true","fetch_graph":"https://pith.science/api/pith-number/UDVBFP75RR6HELUNCKOTVXNLYJ/graph.json","fetch_events":"https://pith.science/api/pith-number/UDVBFP75RR6HELUNCKOTVXNLYJ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/UDVBFP75RR6HELUNCKOTVXNLYJ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/UDVBFP75RR6HELUNCKOTVXNLYJ/action/storage_attestation","attest_author":"https://pith.science/pith/UDVBFP75RR6HELUNCKOTVXNLYJ/action/author_attestation","sign_citation":"https://pith.science/pith/UDVBFP75RR6HELUNCKOTVXNLYJ/action/citation_signature","submit_replication":"https://pith.science/pith/UDVBFP75RR6HELUNCKOTVXNLYJ/action/replication_record"}},"created_at":"2026-05-21T01:04:22.649641+00:00","updated_at":"2026-05-21T01:04:22.649641+00:00"}