{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:QUHPPJQI7GELD5KNU5GIGOEXMS","short_pith_number":"pith:QUHPPJQI","schema_version":"1.0","canonical_sha256":"850ef7a608f988b1f54da74c833897648eb30980c5b69537f8cf51ae559ab28f","source":{"kind":"arxiv","id":"2102.05426","version":2},"attestation_state":"computed","paper":{"title":"BRECQ: Pushing the Limit of Post-Training Quantization by Block Reconstruction","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.CV"],"primary_cat":"cs.LG","authors_text":"Fengwei Yu, Peng Hu, Qi Zhang, Ruihao Gong, Shi Gu, Wei Wang, Xu Tan, Yang Yang, Yuhang Li","submitted_at":"2021-02-10T13:46:16Z","abstract_excerpt":"We study the challenging task of neural network quantization without end-to-end retraining, called Post-training Quantization (PTQ). PTQ usually requires a small subset of training data but produces less powerful quantized models than Quantization-Aware Training (QAT). In this work, we propose a novel PTQ framework, dubbed BRECQ, which pushes the limits of bitwidth in PTQ down to INT2 for the first time. BRECQ leverages the basic building blocks in neural networks and reconstructs them one-by-one. In a comprehensive theoretical study of the second-order error, we show that BRECQ achieves a goo"},"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":"2102.05426","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.LG","submitted_at":"2021-02-10T13:46:16Z","cross_cats_sorted":["cs.CV"],"title_canon_sha256":"ceef1223229744b5bbd0551fe5ac00396a43db3a3119504fa551eb0748e80e4a","abstract_canon_sha256":"3f0010b9947a2ffcc5f8ab7f67cad8f7d15fa38b94c3d754b3ba2ea535da0681"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T03:00:31.656499Z","signature_b64":"Gz/yEWBWVJHRTWhTsNWvDJWMZuBYa3ms8f8VBoFKAaQe/ZOzngxWB4e4BLV+WmMTIegIvaEYKSsipAyujAEWAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"850ef7a608f988b1f54da74c833897648eb30980c5b69537f8cf51ae559ab28f","last_reissued_at":"2026-07-05T03:00:31.656022Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T03:00:31.656022Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"BRECQ: Pushing the Limit of Post-Training Quantization by Block Reconstruction","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.CV"],"primary_cat":"cs.LG","authors_text":"Fengwei Yu, Peng Hu, Qi Zhang, Ruihao Gong, Shi Gu, Wei Wang, Xu Tan, Yang Yang, Yuhang Li","submitted_at":"2021-02-10T13:46:16Z","abstract_excerpt":"We study the challenging task of neural network quantization without end-to-end retraining, called Post-training Quantization (PTQ). PTQ usually requires a small subset of training data but produces less powerful quantized models than Quantization-Aware Training (QAT). In this work, we propose a novel PTQ framework, dubbed BRECQ, which pushes the limits of bitwidth in PTQ down to INT2 for the first time. BRECQ leverages the basic building blocks in neural networks and reconstructs them one-by-one. In a comprehensive theoretical study of the second-order error, we show that BRECQ achieves a goo"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2102.05426","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/2102.05426/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":"2102.05426","created_at":"2026-07-05T03:00:31.656092+00:00"},{"alias_kind":"arxiv_version","alias_value":"2102.05426v2","created_at":"2026-07-05T03:00:31.656092+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2102.05426","created_at":"2026-07-05T03:00:31.656092+00:00"},{"alias_kind":"pith_short_12","alias_value":"QUHPPJQI7GEL","created_at":"2026-07-05T03:00:31.656092+00:00"},{"alias_kind":"pith_short_16","alias_value":"QUHPPJQI7GELD5KN","created_at":"2026-07-05T03:00:31.656092+00:00"},{"alias_kind":"pith_short_8","alias_value":"QUHPPJQI","created_at":"2026-07-05T03:00:31.656092+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":22,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.05429","citing_title":"Minimizing the Hidden Cost of Scales: Graph-Guided Ultra-Low-Bit Quantization for Large Language Models","ref_index":23,"is_internal_anchor":false},{"citing_arxiv_id":"2606.04373","citing_title":"Selective Coupling of Decoupled Informative Regions: Masked Attention Alignment for Data-Free Quantization of Vision Transformers","ref_index":4,"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":138,"is_internal_anchor":false},{"citing_arxiv_id":"2606.01412","citing_title":"GPTQ-intrinsic LoRA: A Near-optimal Algorithm for Low-precision Quantization with Low-rank Adaptation","ref_index":40,"is_internal_anchor":false},{"citing_arxiv_id":"2605.13768","citing_title":"High-Rate Quantized Matrix Multiplication II","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2605.24754","citing_title":"Motion-Compensated Weight Compression","ref_index":32,"is_internal_anchor":false},{"citing_arxiv_id":"2605.26092","citing_title":"GoQuant: Geometric Orthogonal Residual Projection for Multiplier-Free Power-of-Two Transformer Quantization","ref_index":6,"is_internal_anchor":false},{"citing_arxiv_id":"2605.23078","citing_title":"GEMQ: Global Expert-Level Mixed-Precision Quantization for MoE LLMs","ref_index":16,"is_internal_anchor":false},{"citing_arxiv_id":"2408.00923","citing_title":"Reclaiming Residual Knowledge: A Novel Paradigm to Low-Bit Quantization","ref_index":12,"is_internal_anchor":false},{"citing_arxiv_id":"2503.03088","citing_title":"AHCQ-SAM: Toward Accurate and Hardware-Compatible Post-Training Segment Anything Model Quantization","ref_index":17,"is_internal_anchor":false},{"citing_arxiv_id":"2505.02242","citing_title":"Sampling-Aware Quantization for Diffusion Models","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2512.21651","citing_title":"Rethinking Output Alignment For 1-bit Post-Training Quantization of Large Language Models","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2605.16423","citing_title":"Nonlinear Bipolar Compensation: Handling Outliers in Post-Training Quantization","ref_index":17,"is_internal_anchor":false},{"citing_arxiv_id":"2605.17985","citing_title":"SAFE-SVD: Sensitivity-Aware Fidelity-Enforcing SVD for Physics Foundation Models","ref_index":13,"is_internal_anchor":false},{"citing_arxiv_id":"2605.17997","citing_title":"MARR: Module-Adaptive Residual Reconstruction for Low-Bit Post-Training Quantization","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2605.16901","citing_title":"CAR-SAM: Cross-Attention Reconstruction for Post-Training Quantization of the Segment Anything Model","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2512.21651","citing_title":"Rethinking Output Alignment For 1-bit Post-Training Quantization of Large Language Models","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2602.05902","citing_title":"CoreQ: Learning-Free Mismatch Correction and Successive Rounding for Quantization","ref_index":11,"is_internal_anchor":false},{"citing_arxiv_id":"2605.13768","citing_title":"High-Rate Quantized Matrix Multiplication II","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2605.04738","citing_title":"OSAQ: Outlier Self-Absorption for Accurate Low-bit LLM Quantization","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2604.11572","citing_title":"DA-PTQ: Drift-Aware Post-Training Quantization for Efficient Vision-Language-Action Models","ref_index":12,"is_internal_anchor":false},{"citing_arxiv_id":"2605.04738","citing_title":"OSAQ: Outlier Self-Absorption for Accurate Low-bit LLM Quantization","ref_index":10,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/QUHPPJQI7GELD5KNU5GIGOEXMS","json":"https://pith.science/pith/QUHPPJQI7GELD5KNU5GIGOEXMS.json","graph_json":"https://pith.science/api/pith-number/QUHPPJQI7GELD5KNU5GIGOEXMS/graph.json","events_json":"https://pith.science/api/pith-number/QUHPPJQI7GELD5KNU5GIGOEXMS/events.json","paper":"https://pith.science/paper/QUHPPJQI"},"agent_actions":{"view_html":"https://pith.science/pith/QUHPPJQI7GELD5KNU5GIGOEXMS","download_json":"https://pith.science/pith/QUHPPJQI7GELD5KNU5GIGOEXMS.json","view_paper":"https://pith.science/paper/QUHPPJQI","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2102.05426&json=true","fetch_graph":"https://pith.science/api/pith-number/QUHPPJQI7GELD5KNU5GIGOEXMS/graph.json","fetch_events":"https://pith.science/api/pith-number/QUHPPJQI7GELD5KNU5GIGOEXMS/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/QUHPPJQI7GELD5KNU5GIGOEXMS/action/timestamp_anchor","attest_storage":"https://pith.science/pith/QUHPPJQI7GELD5KNU5GIGOEXMS/action/storage_attestation","attest_author":"https://pith.science/pith/QUHPPJQI7GELD5KNU5GIGOEXMS/action/author_attestation","sign_citation":"https://pith.science/pith/QUHPPJQI7GELD5KNU5GIGOEXMS/action/citation_signature","submit_replication":"https://pith.science/pith/QUHPPJQI7GELD5KNU5GIGOEXMS/action/replication_record"}},"created_at":"2026-07-05T03:00:31.656092+00:00","updated_at":"2026-07-05T03:00:31.656092+00:00"}