Research map / Large language models / AI and society
MIME: Minority Inclusion for Majority Group Enhancement of AI Performance
Pradyumna Chari, Yunhao Ba, Shreeram Athreya, Achuta Kadambi
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 10% minority samples to the training set (fixed total size) (ablation) | ResNet-34 gender classifier · FairFace | majority group test accuracy: approximately a 1.5% gain | Improved |
| add minority (dark fur) samples to training (ablation) | ResNet-34 cat/dog classifier · Pets | majority group test accuracy: about 2% | Improved |
| add minority (female) samples to training (ablation) | ResNet-34 age classifier · UTKFace | majority group test accuracy: smaller average improvement at 10% | Improved |
| add minority (female) samples to training (ablation) | ResNet-34 diagnosis classifier · Chest-Xray14 | majority group test accuracy: performance drop for beta = 0.2, prior to an overall gain for beta = 0.3 | Mixed |
| add minority (male) samples to training (ablation) | fully connected income classifier · Adult | majority group test accuracy: prominent accuracy gain for beta = 0.6 | Improved |
| add human (high domain gap) minority samples to chicken majority training (ablation) | ResNet-34 gender classifier · chicken vs human gender (DS-6) | majority group test accuracy: majority accuracy decreases with increasing beta | Hurt |
| keep adding minority samples beyond optimal ratio (ablation) | ResNet-34 · all datasets | majority group test accuracy: lowest accuracy when no majorities are used in training | Hurt |
Closest papers by meaning
- When majority rules, minority loses: bias amplification of gradient descent (NeurIPS 2025 · similarity 0.82)
- Diverse Yet Biased: Towards Mitigating Biases in Generative AI (Student Abstract) (AAAI 2024 · similarity 0.82)
- Boost-and-Skip: A Simple Guidance-Free Diffusion for Minority Generation (ICML 2025 · similarity 0.82)
- The Disparate Benefits of Deep Ensembles (ICML 2025 · similarity 0.82)
- FFB: A Fair Fairness Benchmark for In-Processing Group Fairness Methods (ICLR 2024 · similarity 0.81)
- SKILL-MIX: a Flexible and Expandable Family of Evaluations for AI Models (ICLR 2024 · similarity 0.81)
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.
