AI Paper Atlas

Research map / Architectures and efficiency / Quantization

BRECQ: Pushing the Limit of Post-Training Quantization by Block Reconstruction

Yuhang Li, Ruihao Gong, Xu Tan, Yang Yang, Peng Hu, Qi Zhang, Fengwei Yu, Wei Wang, Shi Gu

ICLR 2021 · Architectures and efficiency › Quantization · approach: Post-training weight quantization · 48 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Use block-wise rather than layer-wise reconstruction (ablation)BRECQ · ImageNet, ResNet-18, 2-bit weightstop-1 accuracy: 65.19 -> 66.39Improved
Use block-wise rather than stage-wise reconstruction (ablation)BRECQ · ImageNet, ResNet-18, 2-bit weightstop-1 accuracy: 66.01 -> 66.39Improved
Use block-wise rather than network-wise reconstruction (ablation)BRECQ · ImageNet, ResNet-18, 2-bit weightstop-1 accuracy: 54.15 -> 66.39Improved
Use block-wise rather than layer-wise reconstruction (ablation)BRECQ · ImageNet, MobileNetV2, 2-bit weightstop-1 accuracy: 52.13 -> 59.67Improved
Use block-wise rather than stage-wise reconstruction (ablation)BRECQ · ImageNet, MobileNetV2, 2-bit weightstop-1 accuracy: 54.23 -> 59.67Improved
Use block-wise rather than network-wise reconstruction (ablation)BRECQ · ImageNet, MobileNetV2, 2-bit weightstop-1 accuracy: 40.76 -> 59.67Improved
Apply 2-bit weight quantization with block reconstructionBRECQ · ImageNet, ResNet-18top-1 accuracy: 55.96 -> 66.30Improved
Apply 2-bit weight quantization with block reconstructionBRECQ · ImageNet, MobileNetV2top-1 accuracy: 32.54 -> 59.67Improved
Apply 2-bit weights and 4-bit activations with block reconstructionBRECQ · ImageNet, ResNet-18top-1 accuracy: 0.21 -> 64.80Improved
Apply 2-bit weights and 4-bit activations with block reconstructionBRECQ · ImageNet, MobileNetV2top-1 accuracy: 0.10 -> 53.34Improved
Use 4-bit post-training quantization instead of QATBRECQ versus PACT · ImageNet, MobileNetV2, 4/4top-1 accuracy: 61.40 -> 66.57Improved
Use 4-bit post-training quantization instead of QATBRECQ versus PACT · ImageNet, ResNet-18, 4/4GPU hours: 100 -> 0.4Improved

Later papers that compared with this method

16 later accepted papers report BRECQ as a baseline in their results tables; 13 of them report a better number on most of the results they share with it.

8 more not listed.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy