{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:OJVTGUDZOXMTOZ5HZE5GWSDNKU","short_pith_number":"pith:OJVTGUDZ","schema_version":"1.0","canonical_sha256":"726b33507975d93767a7c93a6b486d552f8e2871f6b917c11627bf8b76d704ff","source":{"kind":"arxiv","id":"2406.00800","version":2},"attestation_state":"computed","paper":{"title":"MagR: Weight Magnitude Reduction for Enhancing Post-Training Quantization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Aozhong Zhang, Naigang Wang, Penghang Yin, Xin Li, Yanxia Deng, Zi Yang","submitted_at":"2024-06-02T17:00:02Z","abstract_excerpt":"In this paper, we present a simple optimization-based preprocessing technique called Weight Magnitude Reduction (MagR) to improve the performance of post-training quantization. For each linear layer, we adjust the pre-trained floating-point weights by solving an $\\ell_\\infty$-regularized optimization problem. This process greatly diminishes the maximum magnitude of the weights and smooths out outliers, while preserving the layer's output. The preprocessed weights are centered more towards zero, which facilitates the subsequent quantization process. To implement MagR, we address the $\\ell_\\inft"},"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":"2406.00800","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-06-02T17:00:02Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"e51caf049a25511f401e184a3583028c7a716befe41e56aba775b0482bb9a290","abstract_canon_sha256":"ab181a02aa941b0ad3ab68930f51ab1fc33b2da54665633778a81892795c7d1e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:21:44.894815Z","signature_b64":"/owL15+xVoTD5IkUYBbABg+zzFx57DfUcGgH5VeBgUJpLinyTFw20lC+1N+AtEQseSqeiIkubGIAfZIb6Gh+CA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"726b33507975d93767a7c93a6b486d552f8e2871f6b917c11627bf8b76d704ff","last_reissued_at":"2026-07-05T09:21:44.894325Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:21:44.894325Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"MagR: Weight Magnitude Reduction for Enhancing Post-Training Quantization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Aozhong Zhang, Naigang Wang, Penghang Yin, Xin Li, Yanxia Deng, Zi Yang","submitted_at":"2024-06-02T17:00:02Z","abstract_excerpt":"In this paper, we present a simple optimization-based preprocessing technique called Weight Magnitude Reduction (MagR) to improve the performance of post-training quantization. For each linear layer, we adjust the pre-trained floating-point weights by solving an $\\ell_\\infty$-regularized optimization problem. This process greatly diminishes the maximum magnitude of the weights and smooths out outliers, while preserving the layer's output. The preprocessed weights are centered more towards zero, which facilitates the subsequent quantization process. To implement MagR, we address the $\\ell_\\inft"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.00800","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/2406.00800/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":"2406.00800","created_at":"2026-07-05T09:21:44.894384+00:00"},{"alias_kind":"arxiv_version","alias_value":"2406.00800v2","created_at":"2026-07-05T09:21:44.894384+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.00800","created_at":"2026-07-05T09:21:44.894384+00:00"},{"alias_kind":"pith_short_12","alias_value":"OJVTGUDZOXMT","created_at":"2026-07-05T09:21:44.894384+00:00"},{"alias_kind":"pith_short_16","alias_value":"OJVTGUDZOXMTOZ5H","created_at":"2026-07-05T09:21:44.894384+00:00"},{"alias_kind":"pith_short_8","alias_value":"OJVTGUDZ","created_at":"2026-07-05T09:21:44.894384+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.01412","citing_title":"GPTQ-intrinsic LoRA: A Near-optimal Algorithm for Low-precision Quantization with Low-rank Adaptation","ref_index":77,"is_internal_anchor":false},{"citing_arxiv_id":"2508.04853","citing_title":"Provable Post-Training Quantization: Theoretical Analysis of OPTQ and Qronos","ref_index":36,"is_internal_anchor":false},{"citing_arxiv_id":"2604.07674","citing_title":"Weight Group-wise Post-Training Quantization for Medical Foundation Model","ref_index":26,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/OJVTGUDZOXMTOZ5HZE5GWSDNKU","json":"https://pith.science/pith/OJVTGUDZOXMTOZ5HZE5GWSDNKU.json","graph_json":"https://pith.science/api/pith-number/OJVTGUDZOXMTOZ5HZE5GWSDNKU/graph.json","events_json":"https://pith.science/api/pith-number/OJVTGUDZOXMTOZ5HZE5GWSDNKU/events.json","paper":"https://pith.science/paper/OJVTGUDZ"},"agent_actions":{"view_html":"https://pith.science/pith/OJVTGUDZOXMTOZ5HZE5GWSDNKU","download_json":"https://pith.science/pith/OJVTGUDZOXMTOZ5HZE5GWSDNKU.json","view_paper":"https://pith.science/paper/OJVTGUDZ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2406.00800&json=true","fetch_graph":"https://pith.science/api/pith-number/OJVTGUDZOXMTOZ5HZE5GWSDNKU/graph.json","fetch_events":"https://pith.science/api/pith-number/OJVTGUDZOXMTOZ5HZE5GWSDNKU/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/OJVTGUDZOXMTOZ5HZE5GWSDNKU/action/timestamp_anchor","attest_storage":"https://pith.science/pith/OJVTGUDZOXMTOZ5HZE5GWSDNKU/action/storage_attestation","attest_author":"https://pith.science/pith/OJVTGUDZOXMTOZ5HZE5GWSDNKU/action/author_attestation","sign_citation":"https://pith.science/pith/OJVTGUDZOXMTOZ5HZE5GWSDNKU/action/citation_signature","submit_replication":"https://pith.science/pith/OJVTGUDZOXMTOZ5HZE5GWSDNKU/action/replication_record"}},"created_at":"2026-07-05T09:21:44.894384+00:00","updated_at":"2026-07-05T09:21:44.894384+00:00"}