Research map / Architectures and efficiency / Quantization
Distribution-Aware Adaptive Multi-Bit Quantization
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| use DMBQ + LBA instead of ALQ | ResNet-18 · ILSVRC12, weights/activations 2.0/32-bit | Top-1 accuracy: 68.9 -> 70.1 | Helped |
| use DMBQ + LBA instead of ALQ | ResNet-18 · ILSVRC12, weights/activations 2.0/2.0-bit | Top-1 accuracy: 66.4 -> 67.8 | Helped |
| use DMBQ + LBA instead of ALQ | ResNet-34 · ILSVRC12, weights/activations 2.0/2.0-bit | Top-1 accuracy: 71.1 -> 72.1 | Helped |
| use DMBQ + LBA instead of ALQ | ResNet-34 · ILSVRC12, weights/activations 1.0/2.0-bit | Top-1 accuracy: 67.3 -> 69.8 | Helped |
| use DMBQ + LBA instead of LQ-Net | ResNet-20 · CIFAR10, weights/activations 1/2-bit | Top-1 accuracy: 88.4 -> 90.4 | Helped |
| use DMBQ + LBA instead of LQ-Net | VGG-small · CIFAR10, weights/activations 1/2-bit | Top-1 accuracy: 93.4 -> 93.9 | Helped |
| allocate bit-width channel-wise rather than globally or layer-wise (ablation) | VGG-small with DMBQ · CIFAR10, weights 1-bit | Top-1 accuracy: 92.4 -> 93.7 | Helped |
| replace uniform weight quantization with distribution-aware DMBQ (ablation) | ResNet-18 · ILSVRC12, weights/activations 2/2-bit | Top-1 accuracy: 62.7 -> 65.1 | Helped |
| replace uniform weight quantization with distribution-aware DMBQ (ablation) | ResNet-18 · ILSVRC12, weights/activations 3/3-bit | Top-1 accuracy: 68.5 -> 69.2 | Helped |
| replace uniform weight quantization with distribution-aware DMBQ (ablation) | ResNet-18 · ILSVRC12, weights/activations 4/4-bit | Top-1 accuracy: 70.0 -> 70.2 | Helped |
| use DMBQ + LBA instead of ALQ | ResNet-18 training · weights/activations 2.0/32-bit | training time: 2.46× -> 1.16× | Helped |
Closest papers by meaning
- Adaptive Loss-Aware Quantization for Multi-Bit Networks (CVPR 2020 · similarity 0.94)
- Explicit Loss-Error-Aware Quantization for Low-Bit Deep Neural Networks (CVPR 2018 · similarity 0.89)
- Bit-Shrinking: Limiting Instantaneous Sharpness for Improving Post-Training Quantization (CVPR 2023 · similarity 0.89)
- ParetoQ: Improving Scaling Laws in Extremely Low-bit LLM Quantization (NeurIPS 2025 · similarity 0.88)
- Efficient Multi-bit Quantization Network Training via Weight Bias Correction and Bit-wise Coreset Sampling (NeurIPS 2025 · similarity 0.87)
- PTMQ: Post-training Multi-Bit Quantization of Neural Networks (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.