Issues

• ConformalLM with rejection=True computes set loss from unmasked labels, crediting empty and gapped prediction sets

  ml-stat-Sustech/TorchCP#128 · Filter by author shaurya416 opened on Sep 27, 2026

• WSC: covered_train and y_train are passed to calWSC in swapped order, so the worst slice is chosen by class index, not by coverage

  ml-stat-Sustech/TorchCP#127 · Filter by author shaurya416 opened on Sep 26, 2026

• WeightedPredictor: independently sorted weights can change the score-weighted quantile

  ml-stat-Sustech/TorchCP#126 · Filter by author noqt opened on Sep 24, 2026

• LTT p-value in ConformalLM.calibrate_configs: non-integer count passed to binom.cdf, and the factor e dropped for non-binary loss

  ml-stat-Sustech/TorchCP#124 · Filter by author shivamlalakiya opened on Aug 22, 2026

• Can I extract the losses of each method from the

  ml-stat-Sustech/TorchCP#66 · Filter by author marianogabitto opened on Apr 16, 2025
  1

• Evaluating ClusteredPredictor

  ml-stat-Sustech/TorchCP#62 · Filter by author kowshikthopalli opened on Mar 6, 2025
  1

• Locart and Loforest

  ml-stat-Sustech/TorchCP#31 · Filter by author valeman opened on Dec 19, 2024
  1

• Exploiting NN parameters to construct the prediction intervals using ICP

  ml-stat-Sustech/TorchCP#29 · Filter by author Waqar9871 opened on Jul 11, 2024
  1