Research map / Label-efficient and robust learning / Knowledge distillation
Efficient One Pass Self-Distillation with Zipf’s Label Smoothing
Jiajun Liang, Linze Li, Zhaodong Bing, Borui Zhao, Yao Tang, Bo Lin, Haoqiang Fan
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 |
|---|---|---|---|
| set alpha to 1.0 rather than 2.0 (ablation) | Zipf's LS · TinyImageNet | top-1 accuracy: 53.35±0.41 -> 59.25±0.20 | Improved |
| set alpha to 1.0 rather than 0.1 (ablation) | Zipf's LS · CIFAR100 | top-1 accuracy: 77.21±0.29 -> 77.38±0.32 | No clear effect |
| set alpha to 1.0 rather than 0.5 (ablation) | Zipf's LS · TinyImageNet | top-1 accuracy: 59.06±0.21 -> 59.25±0.20 | No clear effect |
| raise lambda from 0.01 to 1.0 (ablation) | Zipf's LS · TinyImageNet | top-1 accuracy: 56.86±0.36 -> 59.25±0.20 | Improved |
| raise lambda from 0.01 to 0.1 (ablation) | Zipf's LS · CIFAR100 | top-1 accuracy: 76.59±0.15 -> 77.38±0.32 | Improved |
| raise lambda from 0.1 to 1.0 (ablation) | Zipf's LS · CIFAR100 | top-1 accuracy: 77.38±0.32 -> 76.79±0.04 | Hurt |
| raise lambda from 1.0 to 1.5 (ablation) | Zipf's LS · TinyImageNet | top-1 accuracy: 59.25±0.20 -> 58.03±0.27 | Hurt |
| increase beta from 0.1 to 0.5 (ablation) | Zipf's LS · CIFAR100 | top-1 accuracy: 77.38±0.32 -> 76.39±0.16 | Hurt |
| increase beta from 0.1 to 0.5 (ablation) | Zipf's LS · TinyImageNet | top-1 accuracy: 58.77±0.18 -> 59.25±0.20 | Mixed |
| replace TF-KD/PS-KD/CE pretrained backbone with Zipf's LS pretrained backbone | ResNet-50 on MS-COCO · MS-COCO detection | AP: 36.4% -> 36.6% | No clear effect |
| use Zipf's LS pretrained backbone instead of TF-KD | ResNet-50 on MS-COCO · MS-COCO detection | AP@0.5: 56.7% -> 58.8% | Improved |
Closest papers by meaning
- Target-Driven Distillation: Consistency Distillation with Target Timestep Selection and Decoupled Guidance (AAAI 2025 · similarity 0.83)
- Rethinking Self-Distillation: Label Averaging and Enhanced Soft Label Refinement with Partial Labels (ICLR 2025 · similarity 0.80)
- Distillation Scaling Laws (ICML 2025 · similarity 0.78)
- Random Conditioning with Distillation for Data-Efficient Diffusion Model Compression (CVPR 2025 · similarity 0.78)
- Score identity Distillation: Exponentially Fast Distillation of Pretrained Diffusion Models for One-Step Generation (ICML 2024 · similarity 0.78)
- Learnable Sampler Distillation for Discrete Diffusion Models (NeurIPS 2025 · similarity 0.77)
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.
