Research map / Architectures and efficiency / Quantization
Accurate Zero-shot Quantization via Hierarchical Teacher-Assistant Distillation
Wonjin Cho, Jeongin Yun, U Kang
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 intermediate-precision teacher assistant (ablation) | Zest · ImageNet-1K val, ResNet-18 W4A4 | top-1 accuracy: 70.23 -> 70.50 | Improved |
| add TA-based image refinement (ablation) | Zest · ImageNet-1K val, ResNet-18 W4A4 | top-1 accuracy: 69.94 -> 70.50 | Improved |
| use block-wise instead of end-to-end reconstruction (ablation) | Zest · ImageNet-1K val, ResNet-18 W4A4 | top-1 accuracy: 69.99 -> 70.50 | Improved |
| use W8A8 TA instead of W16A16 TA (ablation) | Zest · ImageNet-1K val, ResNet-18 W4A4 | top-1 accuracy: 69.15 -> 70.50 | Improved |
| use W8A8 TA instead of W6A6 TA (ablation) | Zest · ImageNet-1K val, ResNet-18 W4A4 | top-1 accuracy: 70.34 -> 70.50 | Improved |
| TA equal to student precision (W4A4) instead of W8A8 (ablation) | Zest · ImageNet-1K val, ResNet-18 W4A4 | top-1 accuracy: 70.50 -> 69.19 | Hurt |
| hierarchical TA distillation vs best baselines at W4A4 | Zest · ImageNet-1K ResNet-50 W4A4 | top-1 accuracy: 75.59 -> 76.32 (Genie) | Improved |
| hierarchical TA distillation vs GenQ at W3A3 | Zest · ImageNet-1K MobileNetV2 W3A3 | top-1 accuracy: 59.15 -> 67.74 (GenQ) | Improved |
| hierarchical TA distillation vs GenQ at W3A3 | Zest · ImageNet-1K ResNet-18 W3A3 | top-1 accuracy: 68.18 -> 68.45 (GenQ) | Improved |
| increase number of TAs beyond one (ablation) | Zest · ImageNet-1K val, ResNet-18 W4A4 | top-1 accuracy | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- GenQ — GenQ: Quantization in Low Data Regimes with Generative Synthetic Data (ECCV 2024) · better on 6 of 6 shared results
- TexQ — TexQ: Zero-shot Network Quantization with Texture Feature Distribution Calibration (NeurIPS 2023) · better on 3 of 3 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
- It's All in the Teacher: Zero-Shot Quantization Brought Closer to the Teacher (CVPR 2022 · similarity 0.93)
- Sharpness-Aware Data Generation for Zero-shot Quantization (ICML 2024 · similarity 0.90)
- SynQ: Accurate Zero-shot Quantization by Synthesis-aware Fine-tuning (ICLR 2025 · similarity 0.89)
- Zero-Shot Quantization via Weight-Space Arithmetic (NeurIPS 2026 · similarity 0.88)
- Quantization without Tears (CVPR 2025 · similarity 0.87)
- 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.
