A masked diffusion model fine-tuned with group-reward reinforcement learning solves Sudoku and Maze puzzles more accurately than the same model trained by supervised learning alone.
During the initial epoch, each puz- zle is sampled 64 times
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.AI 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Constraints-Guided Diffusion Reasoner for Neuro-Symbolic Learning
A masked diffusion model fine-tuned with group-reward reinforcement learning solves Sudoku and Maze puzzles more accurately than the same model trained by supervised learning alone.