Research map / Architectures and efficiency / Quantization
One-Shot Model for Mixed-Precision Quantization
Read the paperCheck 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 |
|---|---|---|---|
| train a bit-width probability model conditioned on the hardware penalty | One-Shot MPS · ImageNet, Top-1 accuracy | Kendall's Tau correlation: 0.52 -> 0.97 | Helped |
| train a bit-width probability model conditioned on the hardware penalty | One-Shot MPS · SRResNet, Set14 | Kendall's Tau correlation: 0.07 -> 0.93 | Helped |
| use Conditional Batch Normalization instead of Batch Normalization (ablation) | One-Shot MPS · SRResNet | Kendall's Tau correlation: 0.52 -> 0.81 | Helped |
| remove Conditional Batch Normalization and use Batch Normalization (ablation) | One-Shot MPS · SRResNet | Kendall's Tau correlation: 0.81 -> 0.52 | Hurt |
| use an exponential temperature schedule or a constant temperature (ablation) | One-Shot MPS · ESPCN, Set14 | PSNR | No clear effect |
| add entropy regularization and initial kernel similarity (ablation) | One-Shot MPS · ESPCN | architecture diversity | Helped |
| remove the entropy term (ablation) | One-Shot MPS · ESPCN | architecture diversity | Hurt |
| use a deeper, two-layer bit-width model instead of a linear model (ablation) | One-Shot MPS · SRResNet | architecture diversity | Helped |
| fit the bit-width model to a wider hardware-penalty range (ablation) | One-Shot MPS · SRResNet, Set14 | PSNR | No clear effect |
| search the Pareto front in one run rather than repeating the search | One-Shot MPS · ResNet-18 and MobileNet-v2 | search time: 5 times | Helped |
Closest papers by meaning
- Retraining-Free Model Quantization via One-Shot Weight-Coupling Learning (CVPR 2024 · similarity 0.92)
- BSQ: Exploring Bit-Level Sparsity for Mixed-Precision Neural Network Quantization (ICLR 2021 · similarity 0.91)
- Effective Interplay between Sparsity and Quantization: From Theory to Practice (ICLR 2025 · similarity 0.89)
- PTMQ: Post-training Multi-Bit Quantization of Neural Networks (AAAI 2024 · similarity 0.88)
- ParetoQ: Improving Scaling Laws in Extremely Low-bit LLM Quantization (NeurIPS 2025 · similarity 0.88)
- Learning from Loss Landscape: Generalizable Mixed-Precision Quantization via Adaptive Sharpness-Aware Gradient Aligning (ICML 2025 · similarity 0.88)
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.