Research map / Architectures and efficiency / Quantization
Dataset Quantization
Daquan Zhou, Kai Wang, Jianyang Gu, Xiangyu Peng, Dongze Lian, Yifan Zhang, Yang You, Jiashi Feng
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 |
|---|---|---|---|
| replace dataset distillation (DM) with dataset quantization for cross-architecture training | CIFAR-10 compressed dataset · CIFAR-10 (rho=10%) | Avg top-1 accuracy: 39.5 -> 69.4 | Improved |
| split dataset into recursively selected bins instead of one-time coreset selection (N=1) (ablation) | DQ · CIFAR-10 | accuracy | Improved |
| use too many bins (N=20) (ablation) | DQ · CIFAR-10 | accuracy | Hurt |
| drop patches by importance score instead of randomly (ablation) | DQ · CIFAR-10 (rho=1%) | accuracy: 41.5 -> 42.3 | Improved |
| increase patch drop ratio (ablation) | DQ · CIFAR-10 | accuracy | Mixed |
| use DQ at 60% data keep ratio for ImageNet pretraining | ResNet-50 on COCO · COCO detection | mAP: 39.2 -> 39.0 | No clear effect |
| instruction-tune LLaMA-7B on 20% DQ-selected Alpaca data instead of full | LLaMA-7B · Alpaca instruction tuning | Avg of BBH, DROP, MMLU, Human-Eval: 27.7 -> 27.1 | No clear effect |
| compress with DQ instead of DM in terms of GPU hours | CIFAR-10 compression · Compression cost | GPU hours: 91 -> 1 | Improved |
| DQ vs random selection at 20-40% keep ratio on downstream tasks | DQ pretraining subset · COCO and ADE20K | mAP / mIoU | Improved |
Closest papers by meaning
- Adaptive Dataset Quantization (AAAI 2025 · similarity 0.90)
- Post Training Quantization for Efficient Dataset Condensation (AAAI 2026 · similarity 0.90)
- Implicit Feature Decoupling With Depthwise Quantization (CVPR 2022 · similarity 0.89)
- Boost Self-Supervised Dataset Distillation via Parameterization, Predefined Augmentation, and Approximation (ICLR 2025 · similarity 0.87)
- Enhancing Diversity for Data-free Quantization (CVPR 2025 · similarity 0.87)
- Large Scale Dataset Distillation with Domain Shift (ICML 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.
