Research map / Architectures and efficiency / Quantization
Leveraging Inter-Layer Dependency for Post -Training Quantization
Changbao Wang, DanDan Zheng, Yuanliu Liu, Liang Li
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 layer-wise activation regularization (AR) (ablation) | NWQ · MobileNetV2 W2A2 | top-1 accuracy: 21.20 -> 26.42 | Improved |
| add layer-wise AR (ablation) | NWQ · MnasNet W2A2 | top-1 accuracy: 30.75 -> 41.17 | Improved |
| add layer-wise AR (ablation) | NWQ · ResNet18 W2A2 | top-1 accuracy: 57.94 -> 59.14 | Improved |
| coarser AR granularity (stage-wise vs layer-wise) (ablation) | NWQ · MobileNetV2 W2A2 | top-1 accuracy: 26.42 -> 25.52 | Mixed |
| block-wise AR instead of layer-wise (ablation) | NWQ · RegNet-3.2GF W2A2 | top-1 accuracy: 62.85 -> 61.31 | Hurt |
| add AutoAugment on top of AR (ablation) | NWQ · MnasNet W2A2 | top-1 accuracy: 41.17 -> 42.92 | Improved |
| replace ASoftmax with RSeR (ablation) | NWQ · MobileNetV2 W2A2 | top-1 accuracy: 26.42 -> 4.86 | Hurt |
| replace ASoftmax with RSeR (ablation) | NWQ · ResNet18 W2A2 | top-1 accuracy: 59.14 -> 46.56 | Hurt |
| use proposed initialization instead of random (ablation) | NWQ · ResNet18 W2A2 | top-1 accuracy: 55.33 -> 59.14 | Improved |
| final temperature tau_T from 1e-2 to 1e-1 (ablation) | ASoftmax · MobileNetV2 W2A2 | top-1 accuracy: 26.42 -> 12.19 | Hurt |
| enlarge discrete range (n=-1, m=2) on small calibration set (ablation) | ASoftmax · RegNet-3.2GF W2A2 | top-1 accuracy: 62.85 -> 61.68 | Hurt |
| scale up calibration set 10x (ablation) | NWQ · MobileNetV2 W2A2 | top-1 accuracy: 26.42 -> 29.54 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- QDROP — QDrop: Randomly Dropping Quantization for Extremely Low-bit Post-Training Quantization (ICLR 2022) · better on 9 of 9 shared results
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
- Optimization-Based Post-Training Quantization With Bit-Split and Stitching (TPAMI 2022 · similarity 0.90)
- BoA: Attention-aware Post-training Quantization without Backpropagation (ICML 2025 · similarity 0.89)
- Pushing the Limit of Post-Training Quantization (TPAMI 2025 · similarity 0.89)
- AIQViT: Architecture-Informed Post-Training Quantization for Vision Transformers (AAAI 2025 · similarity 0.88)
- Enhancing Post-training Quantization Calibration through Contrastive Learning (CVPR 2024 · similarity 0.87)
- Exploring Post-training Quantization in LLMs from Comprehensive Study to Low Rank Compensation (AAAI 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.
