{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:QD3WHZ3EUQHGASKRYWHCSR63E5","short_pith_number":"pith:QD3WHZ3E","schema_version":"1.0","canonical_sha256":"80f763e764a40e604951c58e2947db277bbffbf05c84685abe8d8a881d647f7f","source":{"kind":"arxiv","id":"2306.11987","version":2},"attestation_state":"computed","paper":{"title":"Training Transformers with 4-bit Integers","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.NE"],"primary_cat":"cs.LG","authors_text":"Changhao Li, Haocheng Xi, Jianfei Chen, Jun Zhu","submitted_at":"2023-06-21T02:45:01Z","abstract_excerpt":"Quantizing the activation, weight, and gradient to 4-bit is promising to accelerate neural network training. However, existing 4-bit training methods require custom numerical formats which are not supported by contemporary hardware. In this work, we propose a training method for transformers with all matrix multiplications implemented with the INT4 arithmetic. Training with an ultra-low INT4 precision is challenging. To achieve this, we carefully analyze the specific structures of activation and gradients in transformers to propose dedicated quantizers for them. For forward propagation, we ide"},"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":"2306.11987","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2023-06-21T02:45:01Z","cross_cats_sorted":["cs.NE"],"title_canon_sha256":"4fbda36f20c695ffc211ae220c96a1649d9c3a48b05ab0c411acadec4f1a78ba","abstract_canon_sha256":"4565fafa9371f51b240ec5847d7f3695eaabb3af505ba314f3fbbaf0232083e0"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T06:24:01.140988Z","signature_b64":"0F3Y0tsuR+x5oq6gsvNA+fSLNe32815LVJj3z99RvsozB/r9dbftxQ4aTlFaMHzVUXyG4XPP3hBGc+aNqdb3DQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"80f763e764a40e604951c58e2947db277bbffbf05c84685abe8d8a881d647f7f","last_reissued_at":"2026-07-05T06:24:01.140542Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T06:24:01.140542Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Training Transformers with 4-bit Integers","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.NE"],"primary_cat":"cs.LG","authors_text":"Changhao Li, Haocheng Xi, Jianfei Chen, Jun Zhu","submitted_at":"2023-06-21T02:45:01Z","abstract_excerpt":"Quantizing the activation, weight, and gradient to 4-bit is promising to accelerate neural network training. However, existing 4-bit training methods require custom numerical formats which are not supported by contemporary hardware. In this work, we propose a training method for transformers with all matrix multiplications implemented with the INT4 arithmetic. Training with an ultra-low INT4 precision is challenging. To achieve this, we carefully analyze the specific structures of activation and gradients in transformers to propose dedicated quantizers for them. For forward propagation, we ide"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2306.11987","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/2306.11987/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":"2306.11987","created_at":"2026-07-05T06:24:01.140599+00:00"},{"alias_kind":"arxiv_version","alias_value":"2306.11987v2","created_at":"2026-07-05T06:24:01.140599+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2306.11987","created_at":"2026-07-05T06:24:01.140599+00:00"},{"alias_kind":"pith_short_12","alias_value":"QD3WHZ3EUQHG","created_at":"2026-07-05T06:24:01.140599+00:00"},{"alias_kind":"pith_short_16","alias_value":"QD3WHZ3EUQHGASKR","created_at":"2026-07-05T06:24:01.140599+00:00"},{"alias_kind":"pith_short_8","alias_value":"QD3WHZ3E","created_at":"2026-07-05T06:24:01.140599+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/QD3WHZ3EUQHGASKRYWHCSR63E5","json":"https://pith.science/pith/QD3WHZ3EUQHGASKRYWHCSR63E5.json","graph_json":"https://pith.science/api/pith-number/QD3WHZ3EUQHGASKRYWHCSR63E5/graph.json","events_json":"https://pith.science/api/pith-number/QD3WHZ3EUQHGASKRYWHCSR63E5/events.json","paper":"https://pith.science/paper/QD3WHZ3E"},"agent_actions":{"view_html":"https://pith.science/pith/QD3WHZ3EUQHGASKRYWHCSR63E5","download_json":"https://pith.science/pith/QD3WHZ3EUQHGASKRYWHCSR63E5.json","view_paper":"https://pith.science/paper/QD3WHZ3E","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2306.11987&json=true","fetch_graph":"https://pith.science/api/pith-number/QD3WHZ3EUQHGASKRYWHCSR63E5/graph.json","fetch_events":"https://pith.science/api/pith-number/QD3WHZ3EUQHGASKRYWHCSR63E5/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/QD3WHZ3EUQHGASKRYWHCSR63E5/action/timestamp_anchor","attest_storage":"https://pith.science/pith/QD3WHZ3EUQHGASKRYWHCSR63E5/action/storage_attestation","attest_author":"https://pith.science/pith/QD3WHZ3EUQHGASKRYWHCSR63E5/action/author_attestation","sign_citation":"https://pith.science/pith/QD3WHZ3EUQHGASKRYWHCSR63E5/action/citation_signature","submit_replication":"https://pith.science/pith/QD3WHZ3EUQHGASKRYWHCSR63E5/action/replication_record"}},"created_at":"2026-07-05T06:24:01.140599+00:00","updated_at":"2026-07-05T06:24:01.140599+00:00"}