Research map / Architectures and efficiency / Quantization
REx: Data-Free Residual Quantization Error Expansion
Edouard YVINEC, Arnaud Dapogny, Matthieu Cord, Kevin Bailly
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add REx residual expansion (W4/A6 + 50% sparse residue) vs SPIQ at W6/A6 | MobileNet v2 · ImageNet | top-1 accuracy: 63.24 -> 64.20 | Improved |
| add REx vs SPIQ at W6/A6 | EfficientNet B0 · ImageNet | top-1 accuracy: 54.51 -> 57.63 | Improved |
| add REx vs SPIQ at W6/A6 | ResNet-50 · ImageNet | top-1 accuracy: 75.98 -> 76.01 | Improved |
| add REx residue to DFQ quantization | OPT-13B · common sense reasoning average | average score: 51.47 -> 53.76 | Improved |
| add REx residue to SQuant quantization | OPT-13B · common sense reasoning average | average score: 51.33 -> 53.91 | Improved |
| add REx residue to PowerQuant | OPT-13B · common sense reasoning average | average score: 54.38 -> 54.81 | Improved |
| add REx outlier residue to DFQ | OPT-13B · BoolQ | accuracy: 54.71 -> 65.54 | Improved |
| add REx to PowerQuant | OPT-13B · HellaSwag | accuracy: 51.29 -> 50.98 | Hurt |
| add REx to PowerQuant | OPT-13B · BoolQ | accuracy: 70.43 -> 69.45 | Hurt |
| add REx to DFQ | OPT-13B · OpenBookQA | accuracy: 25.80 -> 25.40 | Hurt |
| add 25% sparse residue to BrecQ in W4/A4 (ablation) | MobileNet v2 · ImageNet | top-1 accuracy: 66.57 -> 70.94 | Improved |
| increase sparse residue from 25% to 75% with SQuant (ablation) | MobileNet v2 · ImageNet W4/A4 | top-1 accuracy: 58.64 -> 71.74 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- SQuant — SQuant: On-the-Fly Data-Free Quantization via Diagonal Hessian Approximation (ICLR 2022) · better on 5 of 5 shared results
- PowerQuant — PowerQuant: Automorphism Search for Non-Uniform Quantization (ICLR 2023) · better on 1 of 1 shared result
- BrecQ — BRECQ: Pushing the Limit of Post-Training Quantization by Block Reconstruction (ICLR 2021) · better on 1 of 1 shared result
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
- Explicit Loss-Error-Aware Quantization for Low-Bit Deep Neural Networks (CVPR 2018 · similarity 0.89)
- Data-Free Quantization Through Weight Equalization and Bias Correction (ICCV 2019 · similarity 0.89)
- Effective Interplay between Sparsity and Quantization: From Theory to Practice (ICLR 2025 · similarity 0.88)
- Quantization without Tears (CVPR 2025 · similarity 0.88)
- Quant Experts: Token-aware Adaptive Error Reconstruction with Mixture of Experts for Large Vision-Language Models Quantization (CVPR 2026 · similarity 0.87)
- Network Memory Footprint Compression Through Jointly Learnable Codebooks and Mappings (ICLR 2024 · similarity 0.87)
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.
