New ProME method enhances group-robust learning without explicit labels

Researchers have introduced ProME (Prototype-Margin Environments), a novel approach to group-robust learning that aims to improve accuracy on rare subpopulations without requiring explicit training-group labels. ProME aligns the selection of representations and the fitting of classifiers with the deployed predictor by splitting prototype margins and using a group-balanced linear head for ranking predictors. This method has demonstrated superior worst-group accuracy compared to existing techniques in extensive experiments. AI

IMPACT Enhances accuracy on rare subpopulations in machine learning models, potentially improving fairness and reliability.

RANK_REASON The cluster contains a research paper detailing a new method for group-robust learning. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• alphaXiv
• arXiv
• CatalyzeX
• DagsHub
• Erasmus University Rotterdam
• Gotit.pub
• group-robust learning
• Hugging Face
• IArxiv
• ScienceCast

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →