Research map / Label-efficient and robust learning / Recognition and long-tail
Easy to Learn, Yet Hard to Forget: Towards Robust Unlearning Under Bias
JuneHyoung Kwon, MiHyeon Kim, Eunju Lee, Yoonji Lee, Seunghoon Lee, YoungBin Kim
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 Sharpness-Aware Partitioning to NegGrad (ablation) | CUPID · Waterbirds (biased train forget set) | FA: 34.96 -> 20.38 | Improved |
| add Sharpness-Aware Partitioning to NegGrad (ablation) | CUPID · Waterbirds (biased train forget set) | RA: 99.81 -> 91.32 | Hurt |
| add Causal Pathway Identification (ablation) | CUPID · Waterbirds (biased train forget set) | FA: 20.38 -> 14.56 | Improved |
| add Causal Pathway Identification (ablation) | CUPID · Waterbirds (biased train forget set) | RA: 91.32 -> 99.45 | Improved |
| add Targeted Pathway Update (ablation) | CUPID · Waterbirds (biased train forget set) | FA: 14.56 -> 6.91 | Improved |
| add projected causal gradient only to the standard gradient (ablation) | CUPID · Waterbirds (unbiased test) | FA: 18.67 -> 12.77 | Improved |
| add bias gradient only to the standard gradient (ablation) | CUPID · Waterbirds (unbiased test) | FA: 18.67 -> 17.73 | No clear effect |
| apply both projected causal and bias gradients to their pathways (ablation) | CUPID · Waterbirds (unbiased test) | FA: 18.67 -> 6.02 | Improved |
| add sharpness reweighting of the update (ablation) | CUPID · Waterbirds (unbiased test) | FA: 17.77 -> 6.02 | Improved |
| add sharpness reweighting of the update (ablation) | CUPID · Waterbirds (unbiased test) | MIA: 30.81 -> 21.79 | Improved |
| choose sharpness percentile k=5% rather than a purer smaller k (ablation) | CUPID · sharpness threshold analysis | performance | Improved |
| compare CUPID with strongest baseline | CUPID vs NegGrad · BAR (unbiased test) | FA: 30.26 -> 3.75 | Improved |
Closest papers by meaning
- When to Forget? Complexity Trade-offs in Machine Unlearning (ICML 2025 · similarity 0.90)
- Data Unlearning in Diffusion Models (ICLR 2025 · similarity 0.89)
- What makes unlearning hard and what to do about it (NeurIPS 2024 · similarity 0.89)
- Efficient Source-free Unlearning via Energy-Guided Data Synthesis and Discrimination-Aware Multitask Optimization (ICML 2025 · similarity 0.89)
- Unlearning or Obfuscating? Jogging the Memory of Unlearned LLMs via Benign Relearning (ICLR 2025 · similarity 0.88)
- Towards Source-Free Machine Unlearning (CVPR 2025 · 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.
