{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:4LOUX63FFR2HW5BKYB3B4VC4O4","short_pith_number":"pith:4LOUX63F","schema_version":"1.0","canonical_sha256":"e2dd4bfb652c747b742ac0761e545c773f7fd85eab76edbbe39f3ea6f8ca1532","source":{"kind":"arxiv","id":"2506.05748","version":1},"attestation_state":"computed","paper":{"title":"Efficient Online RFT with Plug-and-Play LLM Judges: Unlocking State-of-the-Art Performance","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Ananya Pandey, Rudransh Agnihotri","submitted_at":"2025-06-06T05:18:54Z","abstract_excerpt":"Reward-model training is the cost bottleneck in modern Reinforcement Learning Human Feedback (RLHF) pipelines, often requiring tens of billions of parameters and an offline preference-tuning phase. In the proposed method, a frozen, instruction-tuned 7B LLM is augmented with only a one line JSON rubric and a rank-16 LoRA adapter (affecting just 0.8% of the model's parameters), enabling it to serve as a complete substitute for the previously used heavyweight evaluation models. The plug-and-play judge achieves 96.2% accuracy on RewardBench, outperforming specialized reward networks ranging from 2"},"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":"2506.05748","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.LG","submitted_at":"2025-06-06T05:18:54Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"cabb7c70e80ad642f109cdbb06eff01d3438353bc7cc5c6ed386d1c6b957778e","abstract_canon_sha256":"c0941671937150f8892ea3df075659267190e6be84ff8baeb05a203d08279029"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:17:09.030835Z","signature_b64":"PfLvGgRuSoNHqRNo83Y1G09P79oOA1r8K+zPmaoLCRtGvJFcxsHXVo2Vm8MZ2VAnQcd640B0/LUDLFFCmnuRAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e2dd4bfb652c747b742ac0761e545c773f7fd85eab76edbbe39f3ea6f8ca1532","last_reissued_at":"2026-07-05T11:17:09.030228Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:17:09.030228Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Efficient Online RFT with Plug-and-Play LLM Judges: Unlocking State-of-the-Art Performance","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Ananya Pandey, Rudransh Agnihotri","submitted_at":"2025-06-06T05:18:54Z","abstract_excerpt":"Reward-model training is the cost bottleneck in modern Reinforcement Learning Human Feedback (RLHF) pipelines, often requiring tens of billions of parameters and an offline preference-tuning phase. In the proposed method, a frozen, instruction-tuned 7B LLM is augmented with only a one line JSON rubric and a rank-16 LoRA adapter (affecting just 0.8% of the model's parameters), enabling it to serve as a complete substitute for the previously used heavyweight evaluation models. The plug-and-play judge achieves 96.2% accuracy on RewardBench, outperforming specialized reward networks ranging from 2"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.05748","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/2506.05748/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":"2506.05748","created_at":"2026-07-05T11:17:09.030297+00:00"},{"alias_kind":"arxiv_version","alias_value":"2506.05748v1","created_at":"2026-07-05T11:17:09.030297+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.05748","created_at":"2026-07-05T11:17:09.030297+00:00"},{"alias_kind":"pith_short_12","alias_value":"4LOUX63FFR2H","created_at":"2026-07-05T11:17:09.030297+00:00"},{"alias_kind":"pith_short_16","alias_value":"4LOUX63FFR2HW5BK","created_at":"2026-07-05T11:17:09.030297+00:00"},{"alias_kind":"pith_short_8","alias_value":"4LOUX63F","created_at":"2026-07-05T11:17:09.030297+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/4LOUX63FFR2HW5BKYB3B4VC4O4","json":"https://pith.science/pith/4LOUX63FFR2HW5BKYB3B4VC4O4.json","graph_json":"https://pith.science/api/pith-number/4LOUX63FFR2HW5BKYB3B4VC4O4/graph.json","events_json":"https://pith.science/api/pith-number/4LOUX63FFR2HW5BKYB3B4VC4O4/events.json","paper":"https://pith.science/paper/4LOUX63F"},"agent_actions":{"view_html":"https://pith.science/pith/4LOUX63FFR2HW5BKYB3B4VC4O4","download_json":"https://pith.science/pith/4LOUX63FFR2HW5BKYB3B4VC4O4.json","view_paper":"https://pith.science/paper/4LOUX63F","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2506.05748&json=true","fetch_graph":"https://pith.science/api/pith-number/4LOUX63FFR2HW5BKYB3B4VC4O4/graph.json","fetch_events":"https://pith.science/api/pith-number/4LOUX63FFR2HW5BKYB3B4VC4O4/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/4LOUX63FFR2HW5BKYB3B4VC4O4/action/timestamp_anchor","attest_storage":"https://pith.science/pith/4LOUX63FFR2HW5BKYB3B4VC4O4/action/storage_attestation","attest_author":"https://pith.science/pith/4LOUX63FFR2HW5BKYB3B4VC4O4/action/author_attestation","sign_citation":"https://pith.science/pith/4LOUX63FFR2HW5BKYB3B4VC4O4/action/citation_signature","submit_replication":"https://pith.science/pith/4LOUX63FFR2HW5BKYB3B4VC4O4/action/replication_record"}},"created_at":"2026-07-05T11:17:09.030297+00:00","updated_at":"2026-07-05T11:17:09.030297+00:00"}