{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:B6DMMX2ETQV5LCZSEBWAUTNQYE","short_pith_number":"pith:B6DMMX2E","schema_version":"1.0","canonical_sha256":"0f86c65f449c2bd58b32206c0a4db0c1351aa49863abb9ea0ce1bb643b5c1788","source":{"kind":"arxiv","id":"2607.04422","version":1},"attestation_state":"computed","paper":{"title":"Full-Stack FP4: Stable LLM Pretraining with Quantized Projections, Optimizers, and Attention","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Guoqi Li, Jiabo Tong, Mingchuan Ma, Siyu Ding, Xingrun Xing, Ziming Wang","submitted_at":"2026-07-05T17:31:36Z","abstract_excerpt":"Recent NVFP4 pretraining methods mainly target transformer linear layers, leaving optimizer states, optimizer arithmetic and attention underexplored in 4-bit pipelines. This critical gap blocks stable full-stack 4-bit pretraining, as the three core modules exhibit unique numerical failure patterns: linear layers hit hard quantization noise limits with dimension-propagated error amplification; AdamW second moments are heavy-tailed non-negative values fragile to low-precision denominators; attention carries error-prone computation paths demanding strict forward-backward quantization consistency."},"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":"2607.04422","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-07-05T17:31:36Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"7841810ec935ac4bac5d05b37b3ba7acf1b9b0edda4b915b5f5f7ed80308f78d","abstract_canon_sha256":"77df34937e3f82d727afcd49d188f5a17f280959e76d277fe2f08a0282feeca3"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-07T02:19:17.692694Z","signature_b64":"ukKnVahXPV0njWgKR4E7AWOrd+RL9Mom9jPf8Q4Y1d3VKwEoSB2tkItij59o8CZbXEQbHQJjDXwDvxVKXc/HDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0f86c65f449c2bd58b32206c0a4db0c1351aa49863abb9ea0ce1bb643b5c1788","last_reissued_at":"2026-07-07T02:19:17.691942Z","signature_status":"signed_v1","first_computed_at":"2026-07-07T02:19:17.691942Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Full-Stack FP4: Stable LLM Pretraining with Quantized Projections, Optimizers, and Attention","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Guoqi Li, Jiabo Tong, Mingchuan Ma, Siyu Ding, Xingrun Xing, Ziming Wang","submitted_at":"2026-07-05T17:31:36Z","abstract_excerpt":"Recent NVFP4 pretraining methods mainly target transformer linear layers, leaving optimizer states, optimizer arithmetic and attention underexplored in 4-bit pipelines. This critical gap blocks stable full-stack 4-bit pretraining, as the three core modules exhibit unique numerical failure patterns: linear layers hit hard quantization noise limits with dimension-propagated error amplification; AdamW second moments are heavy-tailed non-negative values fragile to low-precision denominators; attention carries error-prone computation paths demanding strict forward-backward quantization consistency."},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2607.04422","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/2607.04422/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":"2607.04422","created_at":"2026-07-07T02:19:17.692050+00:00"},{"alias_kind":"arxiv_version","alias_value":"2607.04422v1","created_at":"2026-07-07T02:19:17.692050+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2607.04422","created_at":"2026-07-07T02:19:17.692050+00:00"},{"alias_kind":"pith_short_12","alias_value":"B6DMMX2ETQV5","created_at":"2026-07-07T02:19:17.692050+00:00"},{"alias_kind":"pith_short_16","alias_value":"B6DMMX2ETQV5LCZS","created_at":"2026-07-07T02:19:17.692050+00:00"},{"alias_kind":"pith_short_8","alias_value":"B6DMMX2E","created_at":"2026-07-07T02:19:17.692050+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/B6DMMX2ETQV5LCZSEBWAUTNQYE","json":"https://pith.science/pith/B6DMMX2ETQV5LCZSEBWAUTNQYE.json","graph_json":"https://pith.science/api/pith-number/B6DMMX2ETQV5LCZSEBWAUTNQYE/graph.json","events_json":"https://pith.science/api/pith-number/B6DMMX2ETQV5LCZSEBWAUTNQYE/events.json","paper":"https://pith.science/paper/B6DMMX2E"},"agent_actions":{"view_html":"https://pith.science/pith/B6DMMX2ETQV5LCZSEBWAUTNQYE","download_json":"https://pith.science/pith/B6DMMX2ETQV5LCZSEBWAUTNQYE.json","view_paper":"https://pith.science/paper/B6DMMX2E","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2607.04422&json=true","fetch_graph":"https://pith.science/api/pith-number/B6DMMX2ETQV5LCZSEBWAUTNQYE/graph.json","fetch_events":"https://pith.science/api/pith-number/B6DMMX2ETQV5LCZSEBWAUTNQYE/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/B6DMMX2ETQV5LCZSEBWAUTNQYE/action/timestamp_anchor","attest_storage":"https://pith.science/pith/B6DMMX2ETQV5LCZSEBWAUTNQYE/action/storage_attestation","attest_author":"https://pith.science/pith/B6DMMX2ETQV5LCZSEBWAUTNQYE/action/author_attestation","sign_citation":"https://pith.science/pith/B6DMMX2ETQV5LCZSEBWAUTNQYE/action/citation_signature","submit_replication":"https://pith.science/pith/B6DMMX2ETQV5LCZSEBWAUTNQYE/action/replication_record"}},"created_at":"2026-07-07T02:19:17.692050+00:00","updated_at":"2026-07-07T02:19:17.692050+00:00"}