{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:NAOOC6B45JEBYY4P6LK6A32XLC","short_pith_number":"pith:NAOOC6B4","schema_version":"1.0","canonical_sha256":"681ce1783cea481c638ff2d5e06f5758a3b84184978da504d3b7d15938ebbfc4","source":{"kind":"arxiv","id":"2509.22536","version":5},"attestation_state":"computed","paper":{"title":"A Comprehensive FP8 Training Recipe for Reasoning-Enhanced Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Congkai Xie, Hongxia Yang, Jiannong Cao, Kejing Yang, Mingfa Feng, Ming Li, Shuo Cai, Wenjun Wang, Yiming Zhang, Zhen Li","submitted_at":"2025-09-26T16:16:49Z","abstract_excerpt":"The immense computational cost of training Large Language Models (LLMs) presents a major barrier to innovation. While FP8 training offers a promising solution with significant theoretical efficiency gains, its widespread adoption has been hindered by the lack of a comprehensive, open-source training recipe. To bridge this gap, we introduce an end-to-end FP8 training recipe that seamlessly integrates continual pre-training and supervised fine-tuning. Our methodology employs a fine-grained, hybrid-granularity quantization strategy to maintain numerical fidelity while maximizing computational eff"},"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":"2509.22536","kind":"arxiv","version":5},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-09-26T16:16:49Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"ab8be7d0f25f03e8235ed7bde2c49b838c0516e17429904a662c77d2848709e6","abstract_canon_sha256":"e0ca84f165577d04a96c1cfcc8c6b96408925e373b59bcf215715c77844b1291"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-08-04T01:57:54.917664Z","signature_b64":"0zB6mvm/GmWndZK7QE5FKBi5WHGJ3HEt7Vc/ZGJRtLyVjSQMzFG0NDKdEBRkR+JcJc2u2KyqMgAfcIoIjDSFDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"681ce1783cea481c638ff2d5e06f5758a3b84184978da504d3b7d15938ebbfc4","last_reissued_at":"2026-08-04T01:57:54.916077Z","signature_status":"signed_v1","first_computed_at":"2026-08-04T01:57:54.916077Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"A Comprehensive FP8 Training Recipe for Reasoning-Enhanced Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Congkai Xie, Hongxia Yang, Jiannong Cao, Kejing Yang, Mingfa Feng, Ming Li, Shuo Cai, Wenjun Wang, Yiming Zhang, Zhen Li","submitted_at":"2025-09-26T16:16:49Z","abstract_excerpt":"The immense computational cost of training Large Language Models (LLMs) presents a major barrier to innovation. While FP8 training offers a promising solution with significant theoretical efficiency gains, its widespread adoption has been hindered by the lack of a comprehensive, open-source training recipe. To bridge this gap, we introduce an end-to-end FP8 training recipe that seamlessly integrates continual pre-training and supervised fine-tuning. Our methodology employs a fine-grained, hybrid-granularity quantization strategy to maintain numerical fidelity while maximizing computational eff"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2509.22536","kind":"arxiv","version":5},"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/2509.22536/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":"2509.22536","created_at":"2026-08-04T01:57:54.917369+00:00"},{"alias_kind":"arxiv_version","alias_value":"2509.22536v5","created_at":"2026-08-04T01:57:54.917369+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2509.22536","created_at":"2026-08-04T01:57:54.917369+00:00"},{"alias_kind":"pith_short_12","alias_value":"NAOOC6B45JEB","created_at":"2026-08-04T01:57:54.917369+00:00"},{"alias_kind":"pith_short_16","alias_value":"NAOOC6B45JEBYY4P","created_at":"2026-08-04T01:57:54.917369+00:00"},{"alias_kind":"pith_short_8","alias_value":"NAOOC6B4","created_at":"2026-08-04T01:57:54.917369+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2605.13030","citing_title":"FeatCal: Feature Calibration for Post-Merging Models","ref_index":65,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/NAOOC6B45JEBYY4P6LK6A32XLC","json":"https://pith.science/pith/NAOOC6B45JEBYY4P6LK6A32XLC.json","graph_json":"https://pith.science/api/pith-number/NAOOC6B45JEBYY4P6LK6A32XLC/graph.json","events_json":"https://pith.science/api/pith-number/NAOOC6B45JEBYY4P6LK6A32XLC/events.json","paper":"https://pith.science/paper/NAOOC6B4"},"agent_actions":{"view_html":"https://pith.science/pith/NAOOC6B45JEBYY4P6LK6A32XLC","download_json":"https://pith.science/pith/NAOOC6B45JEBYY4P6LK6A32XLC.json","view_paper":"https://pith.science/paper/NAOOC6B4","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2509.22536&json=true","fetch_graph":"https://pith.science/api/pith-number/NAOOC6B45JEBYY4P6LK6A32XLC/graph.json","fetch_events":"https://pith.science/api/pith-number/NAOOC6B45JEBYY4P6LK6A32XLC/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/NAOOC6B45JEBYY4P6LK6A32XLC/action/timestamp_anchor","attest_storage":"https://pith.science/pith/NAOOC6B45JEBYY4P6LK6A32XLC/action/storage_attestation","attest_author":"https://pith.science/pith/NAOOC6B45JEBYY4P6LK6A32XLC/action/author_attestation","sign_citation":"https://pith.science/pith/NAOOC6B45JEBYY4P6LK6A32XLC/action/citation_signature","submit_replication":"https://pith.science/pith/NAOOC6B45JEBYY4P6LK6A32XLC/action/replication_record"}},"created_at":"2026-08-04T01:57:54.917369+00:00","updated_at":"2026-08-04T01:57:54.917369+00:00"}