Research map / Label-efficient and robust learning / Recognition and long-tail
Delving into Identify-Emphasize Paradigm for Combating Unknown Bias
Bowen Zhao, Chen Chen, Qian-Wei Wang, Anfeng He, Shu‐Tao Xia
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 self-supervised pretext tasks (ablation) | ECS+GA · Corrupted CIFAR10-1 (rho=99.5%) | unbiased accuracy: 34.3 -> 44.5 | Improved |
| add self-supervised pretext tasks (ablation) | Vanilla · Corrupted CIFAR10-1 (rho=99.5%) | unbiased accuracy: 15.6 -> 23.7 | Improved |
| add confident-picking to epoch-ensemble scoring (ablation) | ECS · Colored MNIST (rho=98%) | average precision: 53.8 -> 95.0 | Improved |
| add peer model (ablation) | ECS · Colored MNIST (rho=98%) | average precision: 95.0 -> 98.8 | Improved |
| add peer model and confident-picking (ablation) | ECS · Biased CelebA | average precision: 46.5 -> 67.6 | Improved |
| add epoch-ensemble to GCE scoring (ablation) | LfF GCE · Colored MNIST (rho=98%) | average precision: 37.0 -> 89.3 | Improved |
| score with ECS instead of vanilla model (ablation) | GA · Colored MNIST (rho=98%) | unbiased accuracy: 49.2 -> 89.5 | Improved |
| train GA on all samples rather than only bias-aligned ones (ablation) | GA · Colored MNIST (rho=98%) | unbiased accuracy: 80.9 -> 89.1 | Improved |
| train GA on all samples rather than only bias-conflicting ones (ablation) | GA · Colored MNIST (rho=98%) | unbiased accuracy: 87.5 -> 89.1 | Improved |
| replace Rew with GA under very few bias-conflicting samples | Rew · Colored MNIST (rho=99.95%) | unbiased accuracy: 22.9 -> 53.9 | Improved |
| evaluate at last epoch instead of best epoch | LfF · Colored MNIST (rho=99%) | unbiased accuracy: 80.3 -> 68.8 | Hurt |
| evaluate at last epoch instead of best epoch | ECS+GA · Colored MNIST (rho=99%) | unbiased accuracy: 86.4 -> 84.6 | Hurt |
Closest papers by meaning
- A Simple Remedy for Dataset Bias via Self-Influence: A Mislabeled Sample Perspective (NeurIPS 2024 · similarity 0.88)
- The Impact of Coreset Selection on Spurious Correlations and Group Robustness (NeurIPS 2025 · similarity 0.87)
- Gradient-Weight Alignment as a Train-Time Proxy for Generalization in Classification Tasks (NeurIPS 2025 · similarity 0.86)
- SEBRA : Debiasing through Self-Guided Bias Ranking (ICLR 2025 · similarity 0.85)
- An Effective Theory of Bias Amplification (ICLR 2025 · similarity 0.85)
- From Hope to Safety: Unlearning Biases of Deep Models via Gradient Penalization in Latent Space (AAAI 2024 · similarity 0.85)
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.
