{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:SW2PVQOYYPLFBPAKMY75FNBCE3","short_pith_number":"pith:SW2PVQOY","schema_version":"1.0","canonical_sha256":"95b4fac1d8c3d650bc0a663fd2b42226e01d52e9cd3fcf04af2a49aee4368b21","source":{"kind":"arxiv","id":"2508.18588","version":1},"attestation_state":"computed","paper":{"title":"History Rhymes: Accelerating LLM Reinforcement Learning with RhymeRL","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.DC"],"primary_cat":"cs.LG","authors_text":"Dong Du, Erhu Feng, Haibo Chen, Jingkai He, Qian Liu, Tao Liu, Tianjian Li, Yubin Xia","submitted_at":"2025-08-26T01:42:46Z","abstract_excerpt":"With the rapid advancement of large language models (LLMs), reinforcement learning (RL) has emerged as a pivotal methodology for enhancing the reasoning capabilities of LLMs. Unlike traditional pre-training approaches, RL encompasses multiple stages: rollout, reward, and training, which necessitates collaboration among various worker types. However, current RL systems continue to grapple with substantial GPU underutilization, due to two primary factors: (1) The rollout stage dominates the overall RL process due to test-time scaling; (2) Imbalances in rollout lengths (within the same batch) res"},"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":"2508.18588","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2025-08-26T01:42:46Z","cross_cats_sorted":["cs.DC"],"title_canon_sha256":"48530c9c413889ec8b04ddcd7fcefcd050b0d06467b469db3c68415a1f4c891e","abstract_canon_sha256":"9d9444e9470d8b8c1ddcfec46bd535f120908772df5290aff4e7de9d7fa9e0ca"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:59:27.252054Z","signature_b64":"y8oc/PR0/RpB/ApG1AMQ8MFOnKUzHT39fcvH1qzOPpmwlyCqkVp/0v6kVT4MWO1xkvYtGnshh8+6+hKrB5LXCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"95b4fac1d8c3d650bc0a663fd2b42226e01d52e9cd3fcf04af2a49aee4368b21","last_reissued_at":"2026-07-05T11:59:27.251636Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:59:27.251636Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"History Rhymes: Accelerating LLM Reinforcement Learning with RhymeRL","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.DC"],"primary_cat":"cs.LG","authors_text":"Dong Du, Erhu Feng, Haibo Chen, Jingkai He, Qian Liu, Tao Liu, Tianjian Li, Yubin Xia","submitted_at":"2025-08-26T01:42:46Z","abstract_excerpt":"With the rapid advancement of large language models (LLMs), reinforcement learning (RL) has emerged as a pivotal methodology for enhancing the reasoning capabilities of LLMs. Unlike traditional pre-training approaches, RL encompasses multiple stages: rollout, reward, and training, which necessitates collaboration among various worker types. However, current RL systems continue to grapple with substantial GPU underutilization, due to two primary factors: (1) The rollout stage dominates the overall RL process due to test-time scaling; (2) Imbalances in rollout lengths (within the same batch) res"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2508.18588","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/2508.18588/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":"2508.18588","created_at":"2026-07-05T11:59:27.251691+00:00"},{"alias_kind":"arxiv_version","alias_value":"2508.18588v1","created_at":"2026-07-05T11:59:27.251691+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2508.18588","created_at":"2026-07-05T11:59:27.251691+00:00"},{"alias_kind":"pith_short_12","alias_value":"SW2PVQOYYPLF","created_at":"2026-07-05T11:59:27.251691+00:00"},{"alias_kind":"pith_short_16","alias_value":"SW2PVQOYYPLFBPAK","created_at":"2026-07-05T11:59:27.251691+00:00"},{"alias_kind":"pith_short_8","alias_value":"SW2PVQOY","created_at":"2026-07-05T11:59:27.251691+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":15,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.18967","citing_title":"EfficientRollout: System-Aware Self-Speculative Decoding for RL Rollouts","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2606.11867","citing_title":"Harnessing Routing Foresight for Micro-step-level MoE load balancing in RL Post-training","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2606.03077","citing_title":"Libra: Efficient Resource Management for Agentic RL Post-Training","ref_index":22,"is_internal_anchor":false},{"citing_arxiv_id":"2605.06534","citing_title":"ROSE: Rollout On Serving GPUs via Cooperative Elasticity for Agentic RL","ref_index":27,"is_internal_anchor":false},{"citing_arxiv_id":"2605.15565","citing_title":"AstraFlow: Dataflow-Oriented Reinforcement Learning for Agentic LLMs","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2605.17570","citing_title":"How Off-Policy Can GRPO Be? Mu-GRPO for Efficient LLM Reinforcement Learning","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2510.19225","citing_title":"RLBoost: Harvesting Preemptible Resources for Cost-Efficient Reinforcement Learning on LLMs","ref_index":16,"is_internal_anchor":false},{"citing_arxiv_id":"2511.14617","citing_title":"Seer: Online Context Learning for Fast Synchronous LLM Reinforcement Learning","ref_index":13,"is_internal_anchor":false},{"citing_arxiv_id":"2605.13907","citing_title":"AIS: Adaptive Importance Sampling for Quantized RL","ref_index":6,"is_internal_anchor":false},{"citing_arxiv_id":"2604.26256","citing_title":"DORA: A Scalable Asynchronous Reinforcement Learning System for Language Model Training","ref_index":6,"is_internal_anchor":false},{"citing_arxiv_id":"2605.08862","citing_title":"BubbleSpec: Turning Long-Tail Bubbles into Speculative Rollout Drafts for Synchronous Reinforcement Learning","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2605.06534","citing_title":"ROSE: Rollout On Serving GPUs via Cooperative Elasticity for Agentic RL","ref_index":27,"is_internal_anchor":false},{"citing_arxiv_id":"2604.09107","citing_title":"TensorHub: Scalable and Elastic Weight Transfer for LLM RL Training","ref_index":11,"is_internal_anchor":false},{"citing_arxiv_id":"2605.07244","citing_title":"Experience Sharing in Mutual Reinforcement Learning for Heterogeneous Language Models","ref_index":28,"is_internal_anchor":false},{"citing_arxiv_id":"2605.06755","citing_title":"Gradient Extrapolation-Based Policy Optimization","ref_index":11,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/SW2PVQOYYPLFBPAKMY75FNBCE3","json":"https://pith.science/pith/SW2PVQOYYPLFBPAKMY75FNBCE3.json","graph_json":"https://pith.science/api/pith-number/SW2PVQOYYPLFBPAKMY75FNBCE3/graph.json","events_json":"https://pith.science/api/pith-number/SW2PVQOYYPLFBPAKMY75FNBCE3/events.json","paper":"https://pith.science/paper/SW2PVQOY"},"agent_actions":{"view_html":"https://pith.science/pith/SW2PVQOYYPLFBPAKMY75FNBCE3","download_json":"https://pith.science/pith/SW2PVQOYYPLFBPAKMY75FNBCE3.json","view_paper":"https://pith.science/paper/SW2PVQOY","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2508.18588&json=true","fetch_graph":"https://pith.science/api/pith-number/SW2PVQOYYPLFBPAKMY75FNBCE3/graph.json","fetch_events":"https://pith.science/api/pith-number/SW2PVQOYYPLFBPAKMY75FNBCE3/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/SW2PVQOYYPLFBPAKMY75FNBCE3/action/timestamp_anchor","attest_storage":"https://pith.science/pith/SW2PVQOYYPLFBPAKMY75FNBCE3/action/storage_attestation","attest_author":"https://pith.science/pith/SW2PVQOYYPLFBPAKMY75FNBCE3/action/author_attestation","sign_citation":"https://pith.science/pith/SW2PVQOYYPLFBPAKMY75FNBCE3/action/citation_signature","submit_replication":"https://pith.science/pith/SW2PVQOYYPLFBPAKMY75FNBCE3/action/replication_record"}},"created_at":"2026-07-05T11:59:27.251691+00:00","updated_at":"2026-07-05T11:59:27.251691+00:00"}