PyTorch implementation of "Train for the Worst, Plan for the Best." Investigating adaptive token ordering in Masked Diffusion Models (MDMs) to sidestep hard subproblems and elicit reasoning in discrete domains.
language-modeling pytorch sudoku-solver reasoning logic-puzzles adaptive-inference machine-learning-research sampling-strategies generative-ai discrete-diffusion masked-diffusion icml-2025 token-ordering discrete-state-space llm-diffusion
-
Updated
Dec 25, 2025