{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2024:FNPD4PXGU4BFCGWCADRUSB7GTJ","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"426827275ed354a9f956802377eb8139613353790a25470518f41f878e208786","cross_cats_sorted":["cs.LG"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-02-27T18:56:19Z","title_canon_sha256":"07386fb3bc4c5e218ecb5dcf1df4250ca8801e5ca786069feec96a2427d82d23"},"schema_version":"1.0","source":{"id":"2402.17764","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2402.17764","created_at":"2026-05-17T23:38:13Z"},{"alias_kind":"arxiv_version","alias_value":"2402.17764v1","created_at":"2026-05-17T23:38:13Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2402.17764","created_at":"2026-05-17T23:38:13Z"},{"alias_kind":"pith_short_12","alias_value":"FNPD4PXGU4BF","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"FNPD4PXGU4BFCGWC","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"FNPD4PXG","created_at":"2026-05-18T12:33:37Z"}],"graph_snapshots":[{"event_id":"sha256:927c57b6765203f0b2100591574c04ffb6c65f90b440e5a707e54cdb1917a590","target":"graph","created_at":"2026-05-17T23:38:13Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"It matches the full-precision (i.e., FP16 or BF16) Transformer LLM with the same model size and training tokens in terms of both perplexity and end-task performance, while being significantly more cost-effective in terms of latency, memory, throughput, and energy consumption."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That the training procedure and scaling law developed for the 1.58-bit ternary setting will continue to produce competitive performance when model size or data volume increases beyond the scales tested."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"BitNet b1.58 shows that ternary 1.58-bit LLMs can match full-precision performance at substantially lower inference cost."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Ternary-weight LLMs achieve full-precision performance at far lower computational cost"}],"snapshot_sha256":"ddc322bdb29f2bfbbc4fa69825a4f1f1d3674a93e814ceed7cf341ae4361abfb"},"formal_canon":{"evidence_count":1,"snapshot_sha256":"c821ee11b34325432ec0696a2de201f332054e83b69091e7efddad8e94353a8e"},"paper":{"abstract_excerpt":"Recent research, such as BitNet, is paving the way for a new era of 1-bit Large Language Models (LLMs). In this work, we introduce a 1-bit LLM variant, namely BitNet b1.58, in which every single parameter (or weight) of the LLM is ternary {-1, 0, 1}. It matches the full-precision (i.e., FP16 or BF16) Transformer LLM with the same model size and training tokens in terms of both perplexity and end-task performance, while being significantly more cost-effective in terms of latency, memory, throughput, and energy consumption. More profoundly, the 1.58-bit LLM defines a new scaling law and recipe f","authors_text":"Furu Wei, Hongyu Wang, Jilong Xue, Lei Wang, Li Dong, Lingxiao Ma, Ruiping Wang, Shaohan Huang, Shuming Ma, Wenhui Wang","cross_cats":["cs.LG"],"headline":"Ternary-weight LLMs achieve full-precision performance at far lower computational cost","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-02-27T18:56:19Z","title":"The Era of 1-bit LLMs: All Large Language Models are in 1.58 Bits"},"references":{"count":15,"internal_anchors":9,"resolved_work":15,"sample":[{"cited_arxiv_id":"1911.11641","doi":"","is_internal_anchor":true,"ref_index":1,"title":"PIQA: Reasoning about Physical Commonsense in Natural Language","work_id":"0d865a62-6376-4606-8d3a-eeb3b6e9ba6d","year":1911},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":2,"title":"arXiv preprint arXiv:2307.13304 , year=","work_id":"bd3fe3b4-ccc3-419b-969e-9a80ded56858","year":null},{"cited_arxiv_id":"1905.10044","doi":"","is_internal_anchor":true,"ref_index":3,"title":"BoolQ: Exploring the Surprising Difficulty of Natural Yes/No Questions","work_id":"511eeb84-4b95-46d5-b14f-50da43f4f19f","year":1905},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":4,"title":"1.1 computing’s energy problem (and what we can do about it)","work_id":"85c69f95-61c6-46d7-a2cd-7076728504f6","year":2014},{"cited_arxiv_id":"2306.00978","doi":"","is_internal_anchor":true,"ref_index":5,"title":"AWQ: Activation-aware Weight Quantization for LLM Compression and Acceleration","work_id":"ea9d1d72-db24-4cae-8c89-4ecd83dd87c1","year":null}],"snapshot_sha256":"e30fcd93a9d7cb80a19c153ea14ecd4c6b57ef294d43ba70ac52e8d0802b8538"},"source":{"id":"2402.17764","kind":"arxiv","version":1},"verdict":{"created_at":"2026-05-17T20:06:27.706475Z","id":"133e1b69-a422-44d1-aadb-0eb10183f4ad","model_set":{"reader":"grok-4.3"},"one_line_summary":"BitNet b1.58 shows that ternary 1.58-bit LLMs can match full-precision performance at substantially lower inference cost.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Ternary-weight LLMs achieve full-precision performance at far lower computational cost","strongest_claim":"It matches the full-precision (i.e., FP16 or BF16) Transformer LLM with the same model size and training tokens in terms of both perplexity and end-task performance, while being significantly more cost-effective in terms of latency, memory, throughput, and energy consumption.","weakest_assumption":"That the training procedure and scaling law developed for the 1.58-bit ternary setting will continue to produce competitive performance when model size or data volume increases beyond the scales tested."}},"verdict_id":"133e1b69-a422-44d1-aadb-0eb10183f4ad"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:36e0df6553e289297bc3e0692343f9df5d37b224ffec02fbacf3664f62cdf025","target":"record","created_at":"2026-05-17T23:38:13Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"426827275ed354a9f956802377eb8139613353790a25470518f41f878e208786","cross_cats_sorted":["cs.LG"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-02-27T18:56:19Z","title_canon_sha256":"07386fb3bc4c5e218ecb5dcf1df4250ca8801e5ca786069feec96a2427d82d23"},"schema_version":"1.0","source":{"id":"2402.17764","kind":"arxiv","version":1}},"canonical_sha256":"2b5e3e3ee6a702511ac200e34907e69a7fedfad8892125210de0673b00108196","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"2b5e3e3ee6a702511ac200e34907e69a7fedfad8892125210de0673b00108196","first_computed_at":"2026-05-17T23:38:13.210836Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-17T23:38:13.210836Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"VY9IE30jHVjpx1rjd2udEpfXgJwjQgDRvRHozFAsvcHCnnzb/fwKJcAozR2hAjRCBhJmbw67xBFzh4QPyzw7Ag==","signature_status":"signed_v1","signed_at":"2026-05-17T23:38:13.211486Z","signed_message":"canonical_sha256_bytes"},"source_id":"2402.17764","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:36e0df6553e289297bc3e0692343f9df5d37b224ffec02fbacf3664f62cdf025","sha256:927c57b6765203f0b2100591574c04ffb6c65f90b440e5a707e54cdb1917a590"],"state_sha256":"0412be1e120047fa256f2238c08c8c6b2a7652d9803dd0d66af15dec593a569b"}