{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:RVO2FDVKSC2IBFVSEY3I6MDCOW","short_pith_number":"pith:RVO2FDVK","schema_version":"1.0","canonical_sha256":"8d5da28eaa90b48096b226368f30627596f9c6b4bab292c77906312463a550a6","source":{"kind":"arxiv","id":"2503.04715","version":7},"attestation_state":"computed","paper":{"title":"Predictable Scale: Part I, Step Law -- Optimal Hyperparameter Scaling Law in Large Language Model Pretraining","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Daxin Jiang, Hanshan Zhang, Haoying Wang, Houyi Li, Ning Ding, Qiufeng Wang, Shijie Xuyang, Shuigeng Zhou, Wenzhen Zheng, Xiangyu Zhang, Yuantao Fan, Zhenyu Ding, Zili Wang","submitted_at":"2025-03-06T18:58:29Z","abstract_excerpt":"The impressive capabilities of Large Language Models (LLMs) across diverse tasks are now well established, yet their effective deployment necessitates careful hyperparameter optimization. Although existing methods have explored the influence of hyperparameters on model performance, a principled and generalizable framework across model architectures and data recipes remains absent. In this study, we conduct an unprecedented empirical investigation training over 3,700 LLMs from scratch across 100 trillion tokens, consuming nearly one million NVIDIA H800 GPU hours to establish a universal Scaling"},"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":"2503.04715","kind":"arxiv","version":7},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2025-03-06T18:58:29Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"a7dba12f2962e8e9d2dea8ac51043baf03853d406a826d2a5d6faab1eb024029","abstract_canon_sha256":"1298849a5e10ac3eaa2d2bcc18406fbfb6243a2dc61fb83235afe79c478ebdb0"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:55:38.653379Z","signature_b64":"2HsqbsqB35cEqnGu0ybAPWvZPOtF03WuUSOJM1q5JCuNxC95aPSJA3lu5LE6b1P/+ZNv0jYQoOOMEll0Ex9vAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"8d5da28eaa90b48096b226368f30627596f9c6b4bab292c77906312463a550a6","last_reissued_at":"2026-07-05T11:55:38.652806Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:55:38.652806Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Predictable Scale: Part I, Step Law -- Optimal Hyperparameter Scaling Law in Large Language Model Pretraining","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Daxin Jiang, Hanshan Zhang, Haoying Wang, Houyi Li, Ning Ding, Qiufeng Wang, Shijie Xuyang, Shuigeng Zhou, Wenzhen Zheng, Xiangyu Zhang, Yuantao Fan, Zhenyu Ding, Zili Wang","submitted_at":"2025-03-06T18:58:29Z","abstract_excerpt":"The impressive capabilities of Large Language Models (LLMs) across diverse tasks are now well established, yet their effective deployment necessitates careful hyperparameter optimization. Although existing methods have explored the influence of hyperparameters on model performance, a principled and generalizable framework across model architectures and data recipes remains absent. In this study, we conduct an unprecedented empirical investigation training over 3,700 LLMs from scratch across 100 trillion tokens, consuming nearly one million NVIDIA H800 GPU hours to establish a universal Scaling"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2503.04715","kind":"arxiv","version":7},"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/2503.04715/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":"2503.04715","created_at":"2026-07-05T11:55:38.652880+00:00"},{"alias_kind":"arxiv_version","alias_value":"2503.04715v7","created_at":"2026-07-05T11:55:38.652880+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2503.04715","created_at":"2026-07-05T11:55:38.652880+00:00"},{"alias_kind":"pith_short_12","alias_value":"RVO2FDVKSC2I","created_at":"2026-07-05T11:55:38.652880+00:00"},{"alias_kind":"pith_short_16","alias_value":"RVO2FDVKSC2IBFVS","created_at":"2026-07-05T11:55:38.652880+00:00"},{"alias_kind":"pith_short_8","alias_value":"RVO2FDVK","created_at":"2026-07-05T11:55:38.652880+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":12,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2607.01487","citing_title":"How to Allocate Your Tokens? Scaling Laws with Training Steps and Batch Size","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2606.15079","citing_title":"Ling and Ring 2.6 Technical Report: Efficient and Instant Agentic Intelligence at Trillion-Parameter Scale","ref_index":65,"is_internal_anchor":false},{"citing_arxiv_id":"2606.05610","citing_title":"Predictable Scaling Laws of Optimal Hyperparameters for LLM Continued Pre-training","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2606.05186","citing_title":"Staged Factorial Screening for Budget-Constrained Micro-Pretraining","ref_index":2,"is_internal_anchor":false},{"citing_arxiv_id":"2606.29158","citing_title":"On the Nonlinearity of Learning Rate Scaling for LLM Training","ref_index":22,"is_internal_anchor":false},{"citing_arxiv_id":"2606.28057","citing_title":"MultiHashFormer: Hash-based Generative Language Models","ref_index":21,"is_internal_anchor":false},{"citing_arxiv_id":"2506.12119","citing_title":"Mixture-of-Experts Can Surpass Dense LLMs Under Strictly Equal Resource","ref_index":24,"is_internal_anchor":false},{"citing_arxiv_id":"2605.21486","citing_title":"Quantifying Hyperparameter Transfer and the Importance of Embedding Layer Learning Rate","ref_index":30,"is_internal_anchor":false},{"citing_arxiv_id":"2605.11518","citing_title":"AutoLLMResearch: Training Research Agents for Automating LLM Experiment Configuration - Learning from Cheap, Optimizing Expensive","ref_index":51,"is_internal_anchor":false},{"citing_arxiv_id":"2605.17000","citing_title":"BoLT: A Benchmark to Democratize Black-box Optimization Research for Expensive LLM Tasks","ref_index":47,"is_internal_anchor":false},{"citing_arxiv_id":"2603.28743","citing_title":"Rethinking Language Model Scaling under Transferable Hypersphere Optimization","ref_index":13,"is_internal_anchor":false},{"citing_arxiv_id":"2605.11518","citing_title":"AutoLLMResearch: Training Research Agents for Automating LLM Experiment Configuration - Learning from Cheap, Optimizing Expensive","ref_index":48,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/RVO2FDVKSC2IBFVSEY3I6MDCOW","json":"https://pith.science/pith/RVO2FDVKSC2IBFVSEY3I6MDCOW.json","graph_json":"https://pith.science/api/pith-number/RVO2FDVKSC2IBFVSEY3I6MDCOW/graph.json","events_json":"https://pith.science/api/pith-number/RVO2FDVKSC2IBFVSEY3I6MDCOW/events.json","paper":"https://pith.science/paper/RVO2FDVK"},"agent_actions":{"view_html":"https://pith.science/pith/RVO2FDVKSC2IBFVSEY3I6MDCOW","download_json":"https://pith.science/pith/RVO2FDVKSC2IBFVSEY3I6MDCOW.json","view_paper":"https://pith.science/paper/RVO2FDVK","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2503.04715&json=true","fetch_graph":"https://pith.science/api/pith-number/RVO2FDVKSC2IBFVSEY3I6MDCOW/graph.json","fetch_events":"https://pith.science/api/pith-number/RVO2FDVKSC2IBFVSEY3I6MDCOW/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/RVO2FDVKSC2IBFVSEY3I6MDCOW/action/timestamp_anchor","attest_storage":"https://pith.science/pith/RVO2FDVKSC2IBFVSEY3I6MDCOW/action/storage_attestation","attest_author":"https://pith.science/pith/RVO2FDVKSC2IBFVSEY3I6MDCOW/action/author_attestation","sign_citation":"https://pith.science/pith/RVO2FDVKSC2IBFVSEY3I6MDCOW/action/citation_signature","submit_replication":"https://pith.science/pith/RVO2FDVKSC2IBFVSEY3I6MDCOW/action/replication_record"}},"created_at":"2026-07-05T11:55:38.652880+00:00","updated_at":"2026-07-05T11:55:38.652880+00:00"}