R3 uses process reward model scores to decide which blocks of a masked diffusion language model's output to remask and regenerate, improving MATH accuracy from 29% to 43% without retraining.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Review, Remask, Refine (R3): Process-Guided Block Diffusion for Text Generation
R3 uses process reward model scores to decide which blocks of a masked diffusion language model's output to remask and regenerate, improving MATH accuracy from 29% to 43% without retraining.