{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2024:7LPQM4CBKI6IJLFXS4SPBULVWL","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"592586c7b486cb0c7461dffbf62ec428773c3a092eac9be768e4c02bd653d800","cross_cats_sorted":["cs.AI","cs.CL","stat.ML"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-05-13T15:50:39Z","title_canon_sha256":"9f4678f3878ae83510a56d06bd2ee031ccdaa6a9660221f34c78267ca1f6e80d"},"schema_version":"1.0","source":{"id":"2405.07863","kind":"arxiv","version":3}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2405.07863","created_at":"2026-05-17T23:38:13Z"},{"alias_kind":"arxiv_version","alias_value":"2405.07863v3","created_at":"2026-05-17T23:38:13Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2405.07863","created_at":"2026-05-17T23:38:13Z"},{"alias_kind":"pith_short_12","alias_value":"7LPQM4CBKI6I","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"7LPQM4CBKI6IJLFX","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"7LPQM4CB","created_at":"2026-05-18T12:33:37Z"}],"graph_snapshots":[{"event_id":"sha256:73971f9f56b5cc332e1cac31b688a3da090ff53653ca356c34a2079e4d3a3459","target":"graph","created_at":"2026-05-17T23:38:13Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"We have shown that supervised fine-tuning (SFT) and iterative RLHF can obtain state-of-the-art performance with fully open-source datasets."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"The proxy preference model built from open-source datasets approximates real human feedback closely enough that online RLHF updates remain beneficial rather than harmful."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"The paper supplies a complete open-source recipe for online iterative RLHF that uses proxy preference models and reaches competitive performance on AlpacaEval-2, Arena-Hard, and MT-Bench."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Online iterative RLHF using proxy preference models from open-source datasets reaches state-of-the-art results on LLM chatbot benchmarks."}],"snapshot_sha256":"e6fa12b1e95f7cd16c20a0d0019b40f432fbbd80191bf69f844eaaae261c014a"},"formal_canon":{"evidence_count":2,"snapshot_sha256":"d12b6b4e4011789984be2d919d3be1f95b12751953a42594f86cb7a4e637d156"},"paper":{"abstract_excerpt":"We present the workflow of Online Iterative Reinforcement Learning from Human Feedback (RLHF) in this technical report, which is widely reported to outperform its offline counterpart by a large margin in the recent large language model (LLM) literature. However, existing open-source RLHF projects are still largely confined to the offline learning setting. In this technical report, we aim to fill in this gap and provide a detailed recipe that is easy to reproduce for online iterative RLHF. In particular, since online human feedback is usually infeasible for open-source communities with limited ","authors_text":"Bo Pang, Caiming Xiong, Doyen Sahoo, HanZe Dong, Han Zhao, Haoxiang Wang, Nan Jiang, Tong Zhang, Wei Xiong, Yingbo Zhou","cross_cats":["cs.AI","cs.CL","stat.ML"],"headline":"Online iterative RLHF using proxy preference models from open-source datasets reaches state-of-the-art results on LLM chatbot benchmarks.","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-05-13T15:50:39Z","title":"RLHF Workflow: From Reward Modeling to Online RLHF"},"references":{"count":23,"internal_anchors":0,"resolved_work":23,"sample":[{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":1,"title":"Flat-leaf parsley is often preferred for its robust flavor","work_id":"3c2694e2-896c-47f0-8e99-d90dd293a9d5","year":null},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":2,"title":"Garlic: Finely chopped or minced raw garlic gives a sharp, pungent taste to the Gremolata","work_id":"795ecb39-7a9d-45e9-87fe-a657ed7c6404","year":null},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":3,"title":"In some variations, you might also find:","work_id":"34e57393-c913-4f41-beef-c1879ce1f78d","year":null},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":4,"title":"Lemon juice: This adds a little extra acidity and freshness to the mix. 5. Fresh basil or mint: These herbs can be used instead of, or in addition to, parsley, depending on the dish and personal prefe","work_id":"ce76a768-aba2-4aaa-b916-bcf98d0042c2","year":2024},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":5,"title":"Squats: Helps to strengthen the lower body, particularly the glutes, quadriceps, and hamstrings. 2. Push-ups: Improves upper body strength, targeting the chest, triceps, and shoulders. 3. Planks: Enga","work_id":"f910d4d1-1e52-4117-91f2-beab3e5ef5b6","year":null}],"snapshot_sha256":"481a0d23db2a484444eb7d32b096bc5398924b4cdb6c5aa0656555238d98c04b"},"source":{"id":"2405.07863","kind":"arxiv","version":3},"verdict":{"created_at":"2026-05-17T16:43:20.020674Z","id":"8c2a5daa-ca4f-4c5d-bf1d-b112b840e968","model_set":{"reader":"grok-4.3"},"one_line_summary":"The paper supplies a complete open-source recipe for online iterative RLHF that uses proxy preference models and reaches competitive performance on AlpacaEval-2, Arena-Hard, and MT-Bench.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Online iterative RLHF using proxy preference models from open-source datasets reaches state-of-the-art results on LLM chatbot benchmarks.","strongest_claim":"We have shown that supervised fine-tuning (SFT) and iterative RLHF can obtain state-of-the-art performance with fully open-source datasets.","weakest_assumption":"The proxy preference model built from open-source datasets approximates real human feedback closely enough that online RLHF updates remain beneficial rather than harmful."}},"verdict_id":"8c2a5daa-ca4f-4c5d-bf1d-b112b840e968"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:4b5cc096351a0133d6e5eccd52435b6ad93db2313d53150cf231ed7f3dc2c0d2","target":"record","created_at":"2026-05-17T23:38:13Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"592586c7b486cb0c7461dffbf62ec428773c3a092eac9be768e4c02bd653d800","cross_cats_sorted":["cs.AI","cs.CL","stat.ML"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-05-13T15:50:39Z","title_canon_sha256":"9f4678f3878ae83510a56d06bd2ee031ccdaa6a9660221f34c78267ca1f6e80d"},"schema_version":"1.0","source":{"id":"2405.07863","kind":"arxiv","version":3}},"canonical_sha256":"fadf067041523c84acb79724f0d175b2cf4b4855d6bbde6f3cae47a8b2a89787","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"fadf067041523c84acb79724f0d175b2cf4b4855d6bbde6f3cae47a8b2a89787","first_computed_at":"2026-05-17T23:38:13.576059Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-17T23:38:13.576059Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"YcKqCIMAlZzWs3QUjs9PomQ9Vt25WerhXj2TU79FE7tPhNK9Gzwb+FfkwqK+Ykbx+NRg9I7x3xn0yLTBAh5PDA==","signature_status":"signed_v1","signed_at":"2026-05-17T23:38:13.576666Z","signed_message":"canonical_sha256_bytes"},"source_id":"2405.07863","source_kind":"arxiv","source_version":3}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:4b5cc096351a0133d6e5eccd52435b6ad93db2313d53150cf231ed7f3dc2c0d2","sha256:73971f9f56b5cc332e1cac31b688a3da090ff53653ca356c34a2079e4d3a3459"],"state_sha256":"8aa3ba1ec7359d326b1cab68fcc75ac7c2d7cb5f664f528fa3428a0b3a4a6166"}