Improved Distribution Estimation in $\ell_\infty$

Collection of new arXiv/stat.ML papers (June 1, 2026) presenting theoretical and algorithmic advances across statistical machine learning: improved minimax and tail bounds for discrete distribution estimation under the ℓ∞ norm; analysis and design principles for Best-of-N preference data and consequences for Bradley–Terry reward learning; comparison (theoretical and empirical) showing last-layer linearization often suffices for epistemic uncertainty quantification in deep nets; minimax lower bounds and near-optimal algorithms for batched stochastic linear bandits with 1-bit-per-batch feedback;