Reinforcement Learning from Human Feedback: A Statistical Perspective

This collection of arXiv ML papers covers RLHF (survey + demo repo), methods for learning from unlabeled/discrete-time data, geometry-aware multi-view embedding (Gromov–Wasserstein), transport/transfer-meta-analysis under covariate shift, variational Bayesian adaptive Kalman filtering for sensor networks with intermittent/corrupted observations, Lipschitz regularity of kernel feature maps, measurement-aware score-based data assimilation, inversion-free natural gradient on Riemannian manifolds, characterization of Gaussian universality breakdown in high-dimensional ERM, and time-warping RNNs (s