{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:JDYNQUMRB6GUIHVTV3QFYENIBK","short_pith_number":"pith:JDYNQUMR","schema_version":"1.0","canonical_sha256":"48f0d851910f8d441eb3aee05c11a80a88a99fc9c39e7189a9a314fa84ffff82","source":{"kind":"arxiv","id":"2506.17828","version":2},"attestation_state":"computed","paper":{"title":"Aligning Frozen LLMs by Reinforcement Learning: An Iterative Reweight-then-Optimize Approach","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Alfredo Garcia, Chenliang Li, Jiaxiang Li, Kaixiang Lin, Mingyi Hong, Siliang Zeng, Songtao Lu, Xinnan Zhang, Zhongruo Wang","submitted_at":"2025-06-21T21:49:02Z","abstract_excerpt":"Aligning large language models (LLMs) with human preferences usually requires fine-tuning methods such as RLHF and DPO. These methods directly optimize the model parameters, so they cannot be used in test-time to improve model performance, nor are they applicable when the model weights are not accessible. In contrast, test-time methods sidestep weight updates by leveraging reward functions to guide and improve output quality. However, they incur high inference costs, and their one-shot guidance is often based on imperfect reward or value functions, leading to suboptimal outputs. In this work, "},"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":"2506.17828","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.LG","submitted_at":"2025-06-21T21:49:02Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"15c0e81bc2106c4a5c11d9abeac7eb15f512b45ece5f9ccd23843970b2a97fab","abstract_canon_sha256":"757127d21a7178737541483338458bd784f94a1c1f9b41e39d9fd9c128812e19"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:31:07.699311Z","signature_b64":"hvi33OErXzGLO9bv60aTGXrywiUitS65wQ5Dj0L2nOHsBV/npxRZs3hJDCMHSrJ0/89YXLr5AI+Rq+lWKJJ0CQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"48f0d851910f8d441eb3aee05c11a80a88a99fc9c39e7189a9a314fa84ffff82","last_reissued_at":"2026-07-05T11:31:07.698840Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:31:07.698840Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Aligning Frozen LLMs by Reinforcement Learning: An Iterative Reweight-then-Optimize Approach","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Alfredo Garcia, Chenliang Li, Jiaxiang Li, Kaixiang Lin, Mingyi Hong, Siliang Zeng, Songtao Lu, Xinnan Zhang, Zhongruo Wang","submitted_at":"2025-06-21T21:49:02Z","abstract_excerpt":"Aligning large language models (LLMs) with human preferences usually requires fine-tuning methods such as RLHF and DPO. These methods directly optimize the model parameters, so they cannot be used in test-time to improve model performance, nor are they applicable when the model weights are not accessible. In contrast, test-time methods sidestep weight updates by leveraging reward functions to guide and improve output quality. However, they incur high inference costs, and their one-shot guidance is often based on imperfect reward or value functions, leading to suboptimal outputs. In this work, "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.17828","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/2506.17828/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":"2506.17828","created_at":"2026-07-05T11:31:07.698903+00:00"},{"alias_kind":"arxiv_version","alias_value":"2506.17828v2","created_at":"2026-07-05T11:31:07.698903+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.17828","created_at":"2026-07-05T11:31:07.698903+00:00"},{"alias_kind":"pith_short_12","alias_value":"JDYNQUMRB6GU","created_at":"2026-07-05T11:31:07.698903+00:00"},{"alias_kind":"pith_short_16","alias_value":"JDYNQUMRB6GUIHVT","created_at":"2026-07-05T11:31:07.698903+00:00"},{"alias_kind":"pith_short_8","alias_value":"JDYNQUMR","created_at":"2026-07-05T11:31:07.698903+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/JDYNQUMRB6GUIHVTV3QFYENIBK","json":"https://pith.science/pith/JDYNQUMRB6GUIHVTV3QFYENIBK.json","graph_json":"https://pith.science/api/pith-number/JDYNQUMRB6GUIHVTV3QFYENIBK/graph.json","events_json":"https://pith.science/api/pith-number/JDYNQUMRB6GUIHVTV3QFYENIBK/events.json","paper":"https://pith.science/paper/JDYNQUMR"},"agent_actions":{"view_html":"https://pith.science/pith/JDYNQUMRB6GUIHVTV3QFYENIBK","download_json":"https://pith.science/pith/JDYNQUMRB6GUIHVTV3QFYENIBK.json","view_paper":"https://pith.science/paper/JDYNQUMR","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2506.17828&json=true","fetch_graph":"https://pith.science/api/pith-number/JDYNQUMRB6GUIHVTV3QFYENIBK/graph.json","fetch_events":"https://pith.science/api/pith-number/JDYNQUMRB6GUIHVTV3QFYENIBK/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/JDYNQUMRB6GUIHVTV3QFYENIBK/action/timestamp_anchor","attest_storage":"https://pith.science/pith/JDYNQUMRB6GUIHVTV3QFYENIBK/action/storage_attestation","attest_author":"https://pith.science/pith/JDYNQUMRB6GUIHVTV3QFYENIBK/action/author_attestation","sign_citation":"https://pith.science/pith/JDYNQUMRB6GUIHVTV3QFYENIBK/action/citation_signature","submit_replication":"https://pith.science/pith/JDYNQUMRB6GUIHVTV3QFYENIBK/action/replication_record"}},"created_at":"2026-07-05T11:31:07.698903+00:00","updated_at":"2026-07-05T11:31:07.698903+00:00"}