Research map / Label-efficient and robust learning / Recognition and long-tail
Fighting Fire with Fire: Contrastive Debiasing without Bias-free Data via Generative Bias-transformation
Yeonsung Jung, Hajin Shim, June Yong Yang, Eunho Yang
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 |
|---|---|---|---|
| Use bias-transformed views with contrastive learning instead of a vanilla classifier | CDvG · Colored MNIST, 0.0% bias-free | accuracy: 12.53 -> 96.16 | Improved |
| Combine CDvG with LfF sample weighting | CDvG · Corrupted CIFAR-10, 0.5% bias-free | accuracy: 31.50 -> 39.07 | Improved |
| Use CDvG rather than the strongest listed baseline | Biased MNIST · Biased MNIST | accuracy: 48.6 -> 49.48 | Improved |
| Use CDvG+LfF rather than LfF | ImageNet-9 · IN-a | accuracy: 29.4 -> 34.6 | Improved |
| Use CDvG+LfF rather than LfF | Waterbirds, ResNet-18 · Waterbirds, worst-group | accuracy: 68.02 -> 74.92 | Improved |
| Add the bias-transformation generator G (ablation) | CDvG on Corrupted CIFAR-10 · Corrupted CIFAR-10, 0.0% bias-free | accuracy: 18.29 -> 28.44 | Improved |
| Add the bias-transformation generator G (ablation) | CDvG on Corrupted CIFAR-10 · Corrupted CIFAR-10, 0.5% bias-free | accuracy: 20.75 -> 31.50 | Improved |
| Add the bias-transformation generator G (ablation) | CDvG on Corrupted CIFAR-10 · Corrupted CIFAR-10, 1.0% bias-free | accuracy: 21.37 -> 33.25 | Improved |
| Add the bias-transformation generator G (ablation) | CDvG on Corrupted CIFAR-10 · Corrupted CIFAR-10, 2.0% bias-free | accuracy: 24.50 -> 35.16 | Improved |
| Add the bias-transformation generator G (ablation) | CDvG on Corrupted CIFAR-10 · Corrupted CIFAR-10, 5.0% bias-free | accuracy: 31.13 -> 42.75 | Improved |
| Add contrastive loss (ablation) | CDvG on Corrupted CIFAR-10 · Corrupted CIFAR-10, 0.0% bias-free | accuracy: 24.25 -> 28.44 | Improved |
| Add contrastive loss (ablation) | CDvG on Corrupted CIFAR-10 · Corrupted CIFAR-10, 0.5% bias-free | accuracy: 26.24 -> 31.50 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- BiaSwap — BiaSwap: Removing Dataset Bias With Bias-Tailored Swapping Augmentation (ICCV 2021) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- BiaSwap: Removing Dataset Bias With Bias-Tailored Swapping Augmentation (ICCV 2021 · similarity 0.89)
- Diffusing DeBias: Synthetic Bias Amplification for Model Debiasing (NeurIPS 2025 · similarity 0.88)
- BendVLM: Test-Time Debiasing of Vision-Language Embeddings (NeurIPS 2024 · similarity 0.87)
- Enhancing Intrinsic Features for Debiasing via Investigating Class-Discerning Common Attributes in Bias-Contrastive Pair (CVPR 2024 · similarity 0.87)
- DeNetDM: Debiasing by Network Depth Modulation (NeurIPS 2024 · similarity 0.87)
- A Unified Debiasing Approach for Vision-Language Models across Modalities and Tasks (NeurIPS 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.
