{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:XKO2JZX77GYTAWPTMQS4KSXCGQ","short_pith_number":"pith:XKO2JZX7","schema_version":"1.0","canonical_sha256":"ba9da4e6fff9b13059f36425c54ae2342775921f5123b67b8abc6f48070a126d","source":{"kind":"arxiv","id":"2505.23247","version":2},"attestation_state":"computed","paper":{"title":"Accelerating RLHF Training with Reward Variance Increase","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","math.OC"],"primary_cat":"cs.LG","authors_text":"Houduo Qi, Yancheng Yuan, Zhexuan Gu, Zonglin Yang","submitted_at":"2025-05-29T08:54:06Z","abstract_excerpt":"Reinforcement learning from human feedback (RLHF) is an essential technique for ensuring that large language models (LLMs) are aligned with human values and preferences during the post-training phase. As an effective RLHF approach, group relative policy optimization (GRPO) has demonstrated success in many LLM-based applications. However, efficient GRPO-based RLHF training remains a challenge. Recent studies reveal that a higher reward variance of the initial policy model leads to faster RLHF training. Inspired by this finding, we propose a practical reward adjustment model to accelerate RLHF t"},"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":"2505.23247","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2025-05-29T08:54:06Z","cross_cats_sorted":["cs.AI","math.OC"],"title_canon_sha256":"831ba3f504c9a999eb8ee32faff11598ddef203b982263fa87c26ab37191494b","abstract_canon_sha256":"e3171dcb21497e680be1e5ae14d160100a2a1c53effd831b1899a112b52c72cc"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:22:40.953108Z","signature_b64":"+/bk+A9AtVUDKxBoYMhpVzYhsA0YQWbKgmbVPgS6cI3ommZK6AKHws1qTGQsAIh29xjDKETH7AiPBipLw0y7CA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"ba9da4e6fff9b13059f36425c54ae2342775921f5123b67b8abc6f48070a126d","last_reissued_at":"2026-07-05T11:22:40.952619Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:22:40.952619Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Accelerating RLHF Training with Reward Variance Increase","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","math.OC"],"primary_cat":"cs.LG","authors_text":"Houduo Qi, Yancheng Yuan, Zhexuan Gu, Zonglin Yang","submitted_at":"2025-05-29T08:54:06Z","abstract_excerpt":"Reinforcement learning from human feedback (RLHF) is an essential technique for ensuring that large language models (LLMs) are aligned with human values and preferences during the post-training phase. As an effective RLHF approach, group relative policy optimization (GRPO) has demonstrated success in many LLM-based applications. However, efficient GRPO-based RLHF training remains a challenge. Recent studies reveal that a higher reward variance of the initial policy model leads to faster RLHF training. Inspired by this finding, we propose a practical reward adjustment model to accelerate RLHF t"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.23247","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/2505.23247/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":"2505.23247","created_at":"2026-07-05T11:22:40.952679+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.23247v2","created_at":"2026-07-05T11:22:40.952679+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.23247","created_at":"2026-07-05T11:22:40.952679+00:00"},{"alias_kind":"pith_short_12","alias_value":"XKO2JZX77GYT","created_at":"2026-07-05T11:22:40.952679+00:00"},{"alias_kind":"pith_short_16","alias_value":"XKO2JZX77GYTAWPT","created_at":"2026-07-05T11:22:40.952679+00:00"},{"alias_kind":"pith_short_8","alias_value":"XKO2JZX7","created_at":"2026-07-05T11:22:40.952679+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/XKO2JZX77GYTAWPTMQS4KSXCGQ","json":"https://pith.science/pith/XKO2JZX77GYTAWPTMQS4KSXCGQ.json","graph_json":"https://pith.science/api/pith-number/XKO2JZX77GYTAWPTMQS4KSXCGQ/graph.json","events_json":"https://pith.science/api/pith-number/XKO2JZX77GYTAWPTMQS4KSXCGQ/events.json","paper":"https://pith.science/paper/XKO2JZX7"},"agent_actions":{"view_html":"https://pith.science/pith/XKO2JZX77GYTAWPTMQS4KSXCGQ","download_json":"https://pith.science/pith/XKO2JZX77GYTAWPTMQS4KSXCGQ.json","view_paper":"https://pith.science/paper/XKO2JZX7","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.23247&json=true","fetch_graph":"https://pith.science/api/pith-number/XKO2JZX77GYTAWPTMQS4KSXCGQ/graph.json","fetch_events":"https://pith.science/api/pith-number/XKO2JZX77GYTAWPTMQS4KSXCGQ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/XKO2JZX77GYTAWPTMQS4KSXCGQ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/XKO2JZX77GYTAWPTMQS4KSXCGQ/action/storage_attestation","attest_author":"https://pith.science/pith/XKO2JZX77GYTAWPTMQS4KSXCGQ/action/author_attestation","sign_citation":"https://pith.science/pith/XKO2JZX77GYTAWPTMQS4KSXCGQ/action/citation_signature","submit_replication":"https://pith.science/pith/XKO2JZX77GYTAWPTMQS4KSXCGQ/action/replication_record"}},"created_at":"2026-07-05T11:22:40.952679+00:00","updated_at":"2026-07-05T11:22:40.952679+00:00"}