Research map / Architectures and efficiency / Quantization
Towards Accurate Low Bit-Width Quantization with Multiple Phase Adaptations
Zhaoyi Yan, Yemin Shi, Yaowei Wang, Mingkui Tan, Zheyang Li, Wenming Tan, Yonghong Tian
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 MPA (multiple phase adaptations) to 3-bit quantization (ablation) | ResNet-18 · ImageNet val | top-1 accuracy (%): 66.9 -> 69.7 | Improved |
| add MPA to 4-bit quantization (ablation) | ResNet-18 · ImageNet val | top-1 accuracy (%): 68.6 -> 70.2 | Improved |
| add MPA to 3-bit quantization (ablation) | ResNet-18 · CIFAR-10 | top-1 accuracy (%): 91.6 -> 93.1 | Improved |
| add MPA to 4-bit quantization (ablation) | ResNet-18 · CIFAR-10 | top-1 accuracy (%): 92.3 -> 93.4 | Improved |
| longer training (64 epochs vs 32 epochs) | MPA ResNet-18 3-bit · ImageNet val | top-1 accuracy (%): 68.9 -> 69.7 | Improved |
| longer training (64 epochs vs 32 epochs) | MPA AlexNet 4-bit · ImageNet val | top-1 accuracy (%): 55.8 -> 56.2 | Improved |
| longer training (64 epochs vs 32 epochs) | MPA VGG-16 3-bit · ImageNet val | top-1 accuracy (%): 70.8 -> 71.4 | Improved |
| longer training (64 epochs vs 32 epochs) | MPA ResNet-50 3-bit · ImageNet val | top-1 accuracy (%): 74.8 -> 75.1 | Improved |
| longer training (64 epochs vs 32 epochs) | MPA ResNet-18 ternary · ImageNet val | top-1 accuracy (%): 67.2 -> 69.2 | Improved |
| linear division of weight centers instead of clustering (ablation) | MPA ResNet-18 3-bit · ImageNet val | top-1 accuracy (%): 69.7 -> 67.6 | Hurt |
| linear division of weight centers instead of clustering (ablation) | MPA ResNet-18 4-bit · ImageNet val | top-1 accuracy (%): 70.2 -> 68.9 | Hurt |
| exponential division of weight centers instead of clustering (ablation) | MPA ResNet-18 3-bit · ImageNet val | top-1 accuracy (%): 69.7 -> 69.5 | No clear effect |
Closest papers by meaning
- AMPA: Adaptive Mixed Precision Allocation for Low-Bit Integer Training (ICML 2024 · similarity 0.87)
- ParetoQ: Improving Scaling Laws in Extremely Low-bit LLM Quantization (NeurIPS 2025 · similarity 0.87)
- PTMQ: Post-training Multi-Bit Quantization of Neural Networks (AAAI 2024 · similarity 0.87)
- Retraining-Free Model Quantization via One-Shot Weight-Coupling Learning (CVPR 2024 · similarity 0.87)
- QBB: Quantization with Binary Bases for LLMs (NeurIPS 2024 · similarity 0.86)
- Efficient Multi-bit Quantization Network Training via Weight Bias Correction and Bit-wise Coreset Sampling (NeurIPS 2025 · similarity 0.86)
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.
