{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:IGKD77JMKTLBJKRNATEYJ3PNPY","short_pith_number":"pith:IGKD77JM","schema_version":"1.0","canonical_sha256":"41943ffd2c54d614aa2d04c984eded7e2c98c83ae4a7c5c26e48c5142baa33ab","source":{"kind":"arxiv","id":"2305.17888","version":1},"attestation_state":"computed","paper":{"title":"LLM-QAT: Data-Free Quantization Aware Training for Large Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Barlas Oguz, Changsheng Zhao, Ernie Chang, Pierre Stock, Raghuraman Krishnamoorthi, Vikas Chandra, Yangyang Shi, Yashar Mehdad, Zechun Liu","submitted_at":"2023-05-29T05:22:11Z","abstract_excerpt":"Several post-training quantization methods have been applied to large language models (LLMs), and have been shown to perform well down to 8-bits. We find that these methods break down at lower bit precision, and investigate quantization aware training for LLMs (LLM-QAT) to push quantization levels even further. We propose a data-free distillation method that leverages generations produced by the pre-trained model, which better preserves the original output distribution and allows quantizing any generative model independent of its training data, similar to post-training quantization methods. In"},"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":"2305.17888","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2023-05-29T05:22:11Z","cross_cats_sorted":[],"title_canon_sha256":"dc054c20085c78aecec0b5a1fd872a65191f4e15724e0bf26c9885882266d0be","abstract_canon_sha256":"333b58c862971ca3a8294fb9a64881b9cf466937650d1aee70e6d96493ce2cd1"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T06:14:50.130554Z","signature_b64":"Z60Pve5O8hT/ReAMAdASmA/cmeZfZskWrgpT9u/9bI5rjqSkmmQS1WHQH/2j3e8JAj34wAFEIp83xaGEu7/uBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"41943ffd2c54d614aa2d04c984eded7e2c98c83ae4a7c5c26e48c5142baa33ab","last_reissued_at":"2026-07-05T06:14:50.130211Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T06:14:50.130211Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"LLM-QAT: Data-Free Quantization Aware Training for Large Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Barlas Oguz, Changsheng Zhao, Ernie Chang, Pierre Stock, Raghuraman Krishnamoorthi, Vikas Chandra, Yangyang Shi, Yashar Mehdad, Zechun Liu","submitted_at":"2023-05-29T05:22:11Z","abstract_excerpt":"Several post-training quantization methods have been applied to large language models (LLMs), and have been shown to perform well down to 8-bits. We find that these methods break down at lower bit precision, and investigate quantization aware training for LLMs (LLM-QAT) to push quantization levels even further. We propose a data-free distillation method that leverages generations produced by the pre-trained model, which better preserves the original output distribution and allows quantizing any generative model independent of its training data, similar to post-training quantization methods. In"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2305.17888","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/2305.17888/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":"2305.17888","created_at":"2026-07-05T06:14:50.130262+00:00"},{"alias_kind":"arxiv_version","alias_value":"2305.17888v1","created_at":"2026-07-05T06:14:50.130262+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2305.17888","created_at":"2026-07-05T06:14:50.130262+00:00"},{"alias_kind":"pith_short_12","alias_value":"IGKD77JMKTLB","created_at":"2026-07-05T06:14:50.130262+00:00"},{"alias_kind":"pith_short_16","alias_value":"IGKD77JMKTLBJKRN","created_at":"2026-07-05T06:14:50.130262+00:00"},{"alias_kind":"pith_short_8","alias_value":"IGKD77JM","created_at":"2026-07-05T06:14:50.130262+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":22,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.24937","citing_title":"The Hitchhiker's Guide to Agentic AI: From Foundations to Systems","ref_index":149,"is_internal_anchor":false},{"citing_arxiv_id":"2606.22249","citing_title":"On the Expressive Power of Weight Quantization in Large Language Models","ref_index":26,"is_internal_anchor":false},{"citing_arxiv_id":"2606.10531","citing_title":"LC-QAT: Data-Efficient 2-Bit QAT for LLMs via Linear-Constrained Vector Quantization","ref_index":11,"is_internal_anchor":false},{"citing_arxiv_id":"2606.10531","citing_title":"LC-QAT: Data-Efficient 2-Bit QAT for LLMs via Linear-Constrained Vector Quantization","ref_index":11,"is_internal_anchor":false},{"citing_arxiv_id":"2606.04945","citing_title":"STaR-Quant: State-Time Consistent Post-Training Quantization for Diffusion Large Language Models","ref_index":135,"is_internal_anchor":false},{"citing_arxiv_id":"2606.04238","citing_title":"Recover-LoRA for Aggressive Quantization: Reclaiming Accuracy in 2-Bit Language Models via Low-Rank Adaptation with Knowledge Distillation on Synthetic Data","ref_index":49,"is_internal_anchor":false},{"citing_arxiv_id":"2606.00539","citing_title":"GNMR: Runtime Stability Control for Low-Precision Large Language Model Training","ref_index":33,"is_internal_anchor":false},{"citing_arxiv_id":"2504.06307","citing_title":"Optimizing Large Language Models: Metrics, Energy Efficiency, and Case Study Insights","ref_index":17,"is_internal_anchor":false},{"citing_arxiv_id":"2505.02242","citing_title":"Sampling-Aware Quantization for Diffusion Models","ref_index":23,"is_internal_anchor":false},{"citing_arxiv_id":"2505.16120","citing_title":"LLM-Powered AI Agent Systems and Their Applications in Industry","ref_index":83,"is_internal_anchor":false},{"citing_arxiv_id":"2605.04062","citing_title":"EdgeRazor: A Lightweight Framework for Large Language Models via Mixed-Precision Quantization-Aware Distillation","ref_index":33,"is_internal_anchor":false},{"citing_arxiv_id":"2508.06974","citing_title":"Rethinking 1-bit Optimization Leveraging Pre-trained Large Language Models","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2602.04163","citing_title":"BPDQ: Bit-Plane Decomposition Quantization on a Variable Grid for Large Language Models","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2307.06435","citing_title":"A Comprehensive Overview of Large Language Models","ref_index":262,"is_internal_anchor":false},{"citing_arxiv_id":"2402.13116","citing_title":"A Survey on Knowledge Distillation of Large Language Models","ref_index":21,"is_internal_anchor":false},{"citing_arxiv_id":"2401.05459","citing_title":"Personal LLM Agents: Insights and Survey about the Capability, Efficiency and Security","ref_index":250,"is_internal_anchor":false},{"citing_arxiv_id":"2605.10655","citing_title":"BCJR-QAT: A Differentiable Relaxation of Trellis-Coded Weight Quantization","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2605.08842","citing_title":"XPERT: Expert Knowledge Transfer for Effective Training of Language Models","ref_index":11,"is_internal_anchor":false},{"citing_arxiv_id":"2604.24088","citing_title":"TACO: Efficient Communication Compression of Intermediate Tensors for Scalable Tensor-Parallel LLM Training","ref_index":34,"is_internal_anchor":false},{"citing_arxiv_id":"2604.22906","citing_title":"Network Edge Inference for Large Language Models: Principles, Techniques, and Opportunities","ref_index":102,"is_internal_anchor":false},{"citing_arxiv_id":"2605.05819","citing_title":"HCInfer: An Efficient Inference System via Error Compensation for Resource-Constrained Devices","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2605.04062","citing_title":"EdgeRazor: A Lightweight Framework for Large Language Models via Mixed-Precision Quantization-Aware Distillation","ref_index":33,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/IGKD77JMKTLBJKRNATEYJ3PNPY","json":"https://pith.science/pith/IGKD77JMKTLBJKRNATEYJ3PNPY.json","graph_json":"https://pith.science/api/pith-number/IGKD77JMKTLBJKRNATEYJ3PNPY/graph.json","events_json":"https://pith.science/api/pith-number/IGKD77JMKTLBJKRNATEYJ3PNPY/events.json","paper":"https://pith.science/paper/IGKD77JM"},"agent_actions":{"view_html":"https://pith.science/pith/IGKD77JMKTLBJKRNATEYJ3PNPY","download_json":"https://pith.science/pith/IGKD77JMKTLBJKRNATEYJ3PNPY.json","view_paper":"https://pith.science/paper/IGKD77JM","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2305.17888&json=true","fetch_graph":"https://pith.science/api/pith-number/IGKD77JMKTLBJKRNATEYJ3PNPY/graph.json","fetch_events":"https://pith.science/api/pith-number/IGKD77JMKTLBJKRNATEYJ3PNPY/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/IGKD77JMKTLBJKRNATEYJ3PNPY/action/timestamp_anchor","attest_storage":"https://pith.science/pith/IGKD77JMKTLBJKRNATEYJ3PNPY/action/storage_attestation","attest_author":"https://pith.science/pith/IGKD77JMKTLBJKRNATEYJ3PNPY/action/author_attestation","sign_citation":"https://pith.science/pith/IGKD77JMKTLBJKRNATEYJ3PNPY/action/citation_signature","submit_replication":"https://pith.science/pith/IGKD77JMKTLBJKRNATEYJ3PNPY/action/replication_record"}},"created_at":"2026-07-05T06:14:50.130262+00:00","updated_at":"2026-07-05T06:14:50.130262+00:00"}