{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:MNNJ2Z7IPKZS3IWUCLIKNWM4R2","short_pith_number":"pith:MNNJ2Z7I","schema_version":"1.0","canonical_sha256":"635a9d67e87ab32da2d412d0a6d99c8e8e98e6bbf7f478ca3c513665a542419c","source":{"kind":"arxiv","id":"2402.10631","version":1},"attestation_state":"computed","paper":{"title":"BitDistiller: Unleashing the Potential of Sub-4-Bit LLMs via Self-Distillation","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Dayou Du, Jiaqi Guo, Ningyi Xu, Shijie Cao, Ting Cao, Xiaowen Chu, Yijia Zhang","submitted_at":"2024-02-16T12:27:15Z","abstract_excerpt":"The upscaling of Large Language Models (LLMs) has yielded impressive advances in natural language processing, yet it also poses significant deployment challenges. Weight quantization has emerged as a widely embraced solution to reduce memory and computational demands. This paper introduces BitDistiller, a framework that synergizes Quantization-Aware Training (QAT) with Knowledge Distillation (KD) to boost the performance of LLMs at ultra-low precisions (sub-4-bit). Specifically, BitDistiller first incorporates a tailored asymmetric quantization and clipping technique to maximally preserve the "},"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":"2402.10631","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-02-16T12:27:15Z","cross_cats_sorted":[],"title_canon_sha256":"b5934fc788053916a125b3279bb3277daafc59449e6885ea162462f23eca8e7c","abstract_canon_sha256":"30ce2e2a80db020d2aca690e0960e256bed4cf00d22c8d8e8fdfe1c295804b31"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:45:58.801787Z","signature_b64":"1XHJAr8/oGUk4KW7iTAGPSaxKc3ar9w4AIK7GzjUBNjX+iwEhkVoQrjaON9zrL7OlEiyTSVKHUwMPMsf/TnLBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"635a9d67e87ab32da2d412d0a6d99c8e8e98e6bbf7f478ca3c513665a542419c","last_reissued_at":"2026-07-05T07:45:58.801363Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:45:58.801363Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"BitDistiller: Unleashing the Potential of Sub-4-Bit LLMs via Self-Distillation","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Dayou Du, Jiaqi Guo, Ningyi Xu, Shijie Cao, Ting Cao, Xiaowen Chu, Yijia Zhang","submitted_at":"2024-02-16T12:27:15Z","abstract_excerpt":"The upscaling of Large Language Models (LLMs) has yielded impressive advances in natural language processing, yet it also poses significant deployment challenges. Weight quantization has emerged as a widely embraced solution to reduce memory and computational demands. This paper introduces BitDistiller, a framework that synergizes Quantization-Aware Training (QAT) with Knowledge Distillation (KD) to boost the performance of LLMs at ultra-low precisions (sub-4-bit). Specifically, BitDistiller first incorporates a tailored asymmetric quantization and clipping technique to maximally preserve the "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2402.10631","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/2402.10631/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":"2402.10631","created_at":"2026-07-05T07:45:58.801446+00:00"},{"alias_kind":"arxiv_version","alias_value":"2402.10631v1","created_at":"2026-07-05T07:45:58.801446+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2402.10631","created_at":"2026-07-05T07:45:58.801446+00:00"},{"alias_kind":"pith_short_12","alias_value":"MNNJ2Z7IPKZS","created_at":"2026-07-05T07:45:58.801446+00:00"},{"alias_kind":"pith_short_16","alias_value":"MNNJ2Z7IPKZS3IWU","created_at":"2026-07-05T07:45:58.801446+00:00"},{"alias_kind":"pith_short_8","alias_value":"MNNJ2Z7I","created_at":"2026-07-05T07:45:58.801446+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2512.06443","citing_title":"Vec-LUT: Vector Table Lookup for Parallel Ultra-Low-Bit LLM Inference on Edge Devices","ref_index":13,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/MNNJ2Z7IPKZS3IWUCLIKNWM4R2","json":"https://pith.science/pith/MNNJ2Z7IPKZS3IWUCLIKNWM4R2.json","graph_json":"https://pith.science/api/pith-number/MNNJ2Z7IPKZS3IWUCLIKNWM4R2/graph.json","events_json":"https://pith.science/api/pith-number/MNNJ2Z7IPKZS3IWUCLIKNWM4R2/events.json","paper":"https://pith.science/paper/MNNJ2Z7I"},"agent_actions":{"view_html":"https://pith.science/pith/MNNJ2Z7IPKZS3IWUCLIKNWM4R2","download_json":"https://pith.science/pith/MNNJ2Z7IPKZS3IWUCLIKNWM4R2.json","view_paper":"https://pith.science/paper/MNNJ2Z7I","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2402.10631&json=true","fetch_graph":"https://pith.science/api/pith-number/MNNJ2Z7IPKZS3IWUCLIKNWM4R2/graph.json","fetch_events":"https://pith.science/api/pith-number/MNNJ2Z7IPKZS3IWUCLIKNWM4R2/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/MNNJ2Z7IPKZS3IWUCLIKNWM4R2/action/timestamp_anchor","attest_storage":"https://pith.science/pith/MNNJ2Z7IPKZS3IWUCLIKNWM4R2/action/storage_attestation","attest_author":"https://pith.science/pith/MNNJ2Z7IPKZS3IWUCLIKNWM4R2/action/author_attestation","sign_citation":"https://pith.science/pith/MNNJ2Z7IPKZS3IWUCLIKNWM4R2/action/citation_signature","submit_replication":"https://pith.science/pith/MNNJ2Z7IPKZS3IWUCLIKNWM4R2/action/replication_record"}},"created_at":"2026-07-05T07:45:58.801446+00:00","updated_at":"2026-07-05T07:45:58.801446+00:00"}