Research map / Label-efficient and robust learning / Recognition and long-tail
From Label Smoothing to Label Relaxation
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 |
|---|---|---|---|
| Replace a single smoothed label distribution with a set of candidate distributions and the LR loss | Image classifiers · MNIST and Fashion-MNIST | ECE: 0.106 ± 0.144 -> 0.007 ± 0.002; 0.155 ± 0.128 -> 0.059 ± 0.008 | Helped |
| Replace label smoothing with label relaxation | VGG16 · CIFAR-10 | accuracy: 0.929 ± 0.001 -> 0.927 ± 0.002 | Hurt |
| Replace label smoothing with label relaxation | VGG16 · CIFAR-100 | ECE: 0.149 ± 0.049 -> 0.133 ± 0.069 | Helped |
| Use label relaxation instead of focal loss | ResNet56 (V2) · CIFAR-10 | ECE: 0.036 ± 0.006 -> 0.059 ± 0.090 | Hurt |
| Use label relaxation instead of focal loss | ResNet56 (V2) · CIFAR-100 | ECE: 0.107 ± 0.017 -> 0.092 ± 0.030 | Helped |
| Replace a single smoothed label distribution with a set-valued target | Image classifiers across the evaluated datasets and models · MNIST, Fashion-MNIST, CIFAR-10, and CIFAR-100 | ECE | Helped |
| Replace precise smoothed targets with relaxed targets | Image classifiers · Image classification datasets | accuracy | Mixed |
| Use label relaxation rather than temperature scaling | Image classifiers · CIFAR-10 and CIFAR-100 | ECE | Mixed |
Closest papers by meaning
- MaxSup: Overcoming Representation Collapse in Label Smoothing (NeurIPS 2025 · similarity 0.89)
- Directional Label Diffusion Model for Learning from Noisy Labels (CVPR 2025 · similarity 0.88)
- Training Deep Neural Networks with Virtual Smoothing Classes (AAAI 2025 · similarity 0.87)
- For Better or For Worse? Learning Minimum Variance Features With Label Augmentation (ICLR 2025 · similarity 0.87)
- Towards Understanding Why Label Smoothing Degrades Selective Classification and How to Fix It (ICLR 2025 · similarity 0.87)
- Does Label Smoothing Help Deep Partial Label Learning? (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.