SOURCE-LINKED INTELLIGENCE

Informed Masking: Structure-Aware Perturbation for Reinforcement Learning in Diffusion Large Language Models

arXiv · AI, language, vision and robotics · article · Sep 22, 2026 · UTC

Diffusion Large Language Models (dLLMs) have emerged as an efficient alternative to autoregressive models, yet aligning them via Reinforcement Learning (RL) requires likelihood surrogates estimated from masked reconstruction subproblems under a small Monte Carlo budget per rollout. Existing methods construct these subproblems by uniform random masking, leaving open the question of which subproblems to prioritize. We identify a systematic upstream/downstream structure in dLLM rollouts. Some tokens, when revealed, trigger large confidence changes in nearby undecoded positions; we call them upstr

Read original source ↗ Open in workspace

recordType

paper

region

Global