arXiv cs.AI / cs.LG / cs.CL·9d agoPrediction-Powered Smoothing and Validation for Disaggregated AI Evaluation#bayesian-models#benchmarks#evaluationAI research