{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:7IMGNHPULBHBE6RY42MQYQBYFH","short_pith_number":"pith:7IMGNHPU","schema_version":"1.0","canonical_sha256":"fa18669df4584e127a38e6990c403829dc3833b81ad3425305255339f3c961b1","source":{"kind":"arxiv","id":"2309.05516","version":5},"attestation_state":"computed","paper":{"title":"Optimize Weight Rounding via Signed Gradient Descent for the Quantization of LLMs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Haihao Shen, Kaokao Lv, Weiwei Zhang, Wenhua Cheng, Xin He, Yi Liu, Yiyang Cai","submitted_at":"2023-09-11T14:58:23Z","abstract_excerpt":"Large Language Models (LLMs) have demonstrated exceptional proficiency in language-related tasks, but their deployment poses significant challenges due to substantial memory and storage requirements. Weight-only quantization has emerged as a promising solution, significantly reducing memory and storage needs without sacrificing too much performance. In this study, we introduce SignRound, a method that leverages signed gradient descent (SignSGD) to optimize rounding values and weight clipping in just 200 steps. SignRound integrates the advantages of Quantization-Aware Training (QAT) and Post-Tr"},"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":"2309.05516","kind":"arxiv","version":5},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2023-09-11T14:58:23Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"f5e686e71c66b30e621de06b0bbd0b46eea3733aa618945934721fbdb87e1906","abstract_canon_sha256":"3f03b19ad7a4e5b9d766f63ab9ff7684eb17923c3efc0f2abefa408f85a07a25"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:17:10.668896Z","signature_b64":"lLi/V7ChJ1qjwqYCT1Y2xoBES7+Km2EqKa+On0B3BicHNVm1uvdjbEPJLn4pgbbGswitDag8tNt/44EWx0QjBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"fa18669df4584e127a38e6990c403829dc3833b81ad3425305255339f3c961b1","last_reissued_at":"2026-07-05T09:17:10.668377Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:17:10.668377Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Optimize Weight Rounding via Signed Gradient Descent for the Quantization of LLMs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Haihao Shen, Kaokao Lv, Weiwei Zhang, Wenhua Cheng, Xin He, Yi Liu, Yiyang Cai","submitted_at":"2023-09-11T14:58:23Z","abstract_excerpt":"Large Language Models (LLMs) have demonstrated exceptional proficiency in language-related tasks, but their deployment poses significant challenges due to substantial memory and storage requirements. Weight-only quantization has emerged as a promising solution, significantly reducing memory and storage needs without sacrificing too much performance. In this study, we introduce SignRound, a method that leverages signed gradient descent (SignSGD) to optimize rounding values and weight clipping in just 200 steps. SignRound integrates the advantages of Quantization-Aware Training (QAT) and Post-Tr"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2309.05516","kind":"arxiv","version":5},"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/2309.05516/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":"2309.05516","created_at":"2026-07-05T09:17:10.668448+00:00"},{"alias_kind":"arxiv_version","alias_value":"2309.05516v5","created_at":"2026-07-05T09:17:10.668448+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2309.05516","created_at":"2026-07-05T09:17:10.668448+00:00"},{"alias_kind":"pith_short_12","alias_value":"7IMGNHPULBHB","created_at":"2026-07-05T09:17:10.668448+00:00"},{"alias_kind":"pith_short_16","alias_value":"7IMGNHPULBHBE6RY","created_at":"2026-07-05T09:17:10.668448+00:00"},{"alias_kind":"pith_short_8","alias_value":"7IMGNHPU","created_at":"2026-07-05T09:17:10.668448+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":6,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.23419","citing_title":"GRINQH: Graded Input-based Quantization Hierarchy for Efficient LLM Generation","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2605.14844","citing_title":"XFP: Quality-Targeted Adaptive Codebook Quantization with Sparse Outlier Separation for LLM Inference","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2605.25203","citing_title":"Influence-Inspired Spectral Rotations for Extreme Low-Bit LLM Quantization","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2605.26339","citing_title":"QAM-W: Joint 2D Codebook Quantization for LLM Weights via Hadamard Rotation and Activation-Aware Scaling","ref_index":1,"is_internal_anchor":false},{"citing_arxiv_id":"2502.15761","citing_title":"AIvaluateXR: An Evaluation Framework for on-Device AI in XR with Benchmarking Results","ref_index":43,"is_internal_anchor":false},{"citing_arxiv_id":"2605.01637","citing_title":"The Banach-Butterfly Invariant: Influence-Adaptive Walsh Geometry for Ternary Polynomial Threshold Functions","ref_index":17,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/7IMGNHPULBHBE6RY42MQYQBYFH","json":"https://pith.science/pith/7IMGNHPULBHBE6RY42MQYQBYFH.json","graph_json":"https://pith.science/api/pith-number/7IMGNHPULBHBE6RY42MQYQBYFH/graph.json","events_json":"https://pith.science/api/pith-number/7IMGNHPULBHBE6RY42MQYQBYFH/events.json","paper":"https://pith.science/paper/7IMGNHPU"},"agent_actions":{"view_html":"https://pith.science/pith/7IMGNHPULBHBE6RY42MQYQBYFH","download_json":"https://pith.science/pith/7IMGNHPULBHBE6RY42MQYQBYFH.json","view_paper":"https://pith.science/paper/7IMGNHPU","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2309.05516&json=true","fetch_graph":"https://pith.science/api/pith-number/7IMGNHPULBHBE6RY42MQYQBYFH/graph.json","fetch_events":"https://pith.science/api/pith-number/7IMGNHPULBHBE6RY42MQYQBYFH/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/7IMGNHPULBHBE6RY42MQYQBYFH/action/timestamp_anchor","attest_storage":"https://pith.science/pith/7IMGNHPULBHBE6RY42MQYQBYFH/action/storage_attestation","attest_author":"https://pith.science/pith/7IMGNHPULBHBE6RY42MQYQBYFH/action/author_attestation","sign_citation":"https://pith.science/pith/7IMGNHPULBHBE6RY42MQYQBYFH/action/citation_signature","submit_replication":"https://pith.science/pith/7IMGNHPULBHBE6RY42MQYQBYFH/action/replication_record"}},"created_at":"2026-07-05T09:17:10.668448+00:00","updated_at":"2026-07-05T09:17:10.668448+00:00"}