Research map / Federated learning and privacy / Differential privacy
Label Smoothing Improves Machine Unlearning
Zonglin Di, Zhaowei Zhu, Jinghan Jia, Jiancheng Liu, Zafar Takhirov, Bo Jiang, Yuanshun Yao, Sijia Liu, Yang Liu
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 negative label smoothing to gradient ascent (UGradSL) | gradient ascent · CIFAR-10 class-wise | UA: 25.19 -> 94.99 | Improved |
| add gradient-mixed plus NLS to fine-tuning (UGradSL+) | fine-tuning · CIFAR-100 random forgetting | UA: 2.55 -> 63.89 | Improved |
| add NLS to gradient-mixed UGradSL (ablation) | UGradSL · CIFAR-10 random forgetting | UA: 0.58 -> 20.77 | Improved |
| add gradient-mixed to GA with NLS (ablation) | UGradSL · CIFAR-10 random forgetting | RA / TA: RA 76.41 -> 79.83, TA 70.15 -> 73.94 | Improved |
| use negative instead of positive label smoothing (ablation) | UGradSL · CIFAR-10 random forgetting | UA: 2.02 -> 20.77 | Improved |
| use negative instead of positive label smoothing (ablation) | UGradSL+ · CIFAR-10 random forgetting | UA: 10.81 -> 25.13 | Improved |
| add NLS to UGradSL+ gradient-mixed (ablation) | UGradSL+ · CIFAR-10 random forgetting | UA: 14.12 -> 25.13 | Improved |
| increase forgetting strength in random forgetting | UGradSL+ · CIFAR-100 random forgetting | TA: 70.51 (retrain) vs 61.09 | Hurt |
| tune UGradSL+ toward retrain gap objective (R) | UGradSL+ · CIFAR-100 random forgetting | Avg. Gap: 17.40 -> 3.75 | Improved |
| use UGradSL+ vs GA at scale | UGradSL+ · ImageNet class-wise | UA: 81.23 -> 100.00 | Improved |
| use UGradSL (cheap) vs FT-based UGradSL+ | UGradSL · CIFAR-10 class-wise | RTE (min): 3.07 -> 0.22 | Improved |
| apply UGradSL+ to 20 Newsgroup text | UGradSL+ · 20 Newsgroup random forgetting | UA: 2.26 -> 35.53 | Improved |
Closest papers by meaning
- When to Forget? Complexity Trade-offs in Machine Unlearning (ICML 2025 · similarity 0.88)
- Certified Unlearning for Neural Networks (ICML 2025 · similarity 0.88)
- The Utility and Complexity of In- and Out-of-Distribution Machine Unlearning (ICLR 2025 · similarity 0.88)
- Langevin Unlearning: A New Perspective of Noisy Gradient Descent for Machine Unlearning (NeurIPS 2024 · similarity 0.88)
- Leveraging Per-Instance Privacy for Machine Unlearning (ICML 2025 · similarity 0.88)
- Certified Machine Unlearning via Noisy Stochastic Gradient Descent (NeurIPS 2024 · 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.
