Research map / Architectures and efficiency / Pruning and efficient tuning
Intra-Processing Methods for Debiasing Neural Networks
Yash Savani, Colin White, Naveen Sundar Govindarajulu
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 intra-processing instead of post-processing | Fine-tuned neural networks · Four evaluated datasets | objective score under a bias constraint | Improved |
| Use adversarial fine-tuning rather than random perturbation or layerwise optimization | CelebA ResNet model · CelebA | balanced accuracy under the fairness constraint: 0.914; 0.917; 0.905 | Improved |
| Train separate models with different random seeds | Neural network · ACI (gender), BM (age), COMPAS (race) | standard deviation of bias versus accuracy: 0.012; 0.017; 0.011; 0.002; 0.027; 0.051; 0.015; 0.002; 0.017; 0.027; 0.016; 0.006 | Mixed |
| Use a neural network instead of logistic regression or random forest | Tabular classification · ACI | accuracy: 0.852 $\pm$ 0.000; 0.855 $\pm$ 0.002; 0.844 $\pm$ 0.002 | Improved |
| Use a neural network instead of logistic regression or random forest | Tabular classification · BM | accuracy: 0.901 $\pm$ 0.000; 0.901 $\pm$ 0.002; 0.899 $\pm$ 0.001 | No clear effect |
| Use a neural network instead of logistic regression or random forest | Tabular classification · COMPAS | accuracy: 0.677 $\pm$ 0.000; 0.641 $\pm$ 0.061; 0.652 $\pm$ 0.006 | Hurt |
Closest papers by meaning
- Fast Model DeBias with Machine Unlearning (NeurIPS 2023 · similarity 0.82)
- Diffusing DeBias: Synthetic Bias Amplification for Model Debiasing (NeurIPS 2025 · similarity 0.80)
- Limits to scalable evaluation at the frontier: LLM as judge won’t beat twice the data (ICLR 2025 · similarity 0.78)
- A Simple Remedy for Dataset Bias via Self-Influence: A Mislabeled Sample Perspective (NeurIPS 2024 · similarity 0.78)
- Generalizing Orthogonalization for Models with Non-Linearities (ICML 2024 · similarity 0.78)
- Self-supervised Debiasing Using Low Rank Regularization (CVPR 2024 · similarity 0.77)
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.
