Newest papers

What Makes a Good Unmasking Policy? A Controlled Study of Learned Unmasking in Diffusion Language Models

Jinya Sakurai, Xun Xu, Jaehong Yoon · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.31576 · cited by 0 works (OpenAlex) · open-access PDF

Japanese

マスクされた変換言語モデルは並行してトークンを生成しますが、その効率はどのような予測が執行されるかによって異なります。学習された解除マスクのポリシーはスピード - 正確性の交換を向上させますが、監督、入力、および容量は異なりますが、どの選択が重要であるかを遮断します。私たちは監督されたポリシーを教師を通じて解読手順を分散させ、教師が何を義務としてラベル化しているかを独立して調べますが、ポリシーがそれを模し、ポリシーが何を観察しているか、そして機能評価の数(NFE)がどの程度減らなければならないかを遮断します。我々は、最終出力に匹敵するラベル化の予測が

Machine translation (Meta M2M100); the English original follows.

Abstract

Masked diffusion language models generate tokens in parallel, but their efficiency depends on which predictions are committed. Learned unmasking policies improve the speed--accuracy trade-off, yet differ in supervision, inputs, and capacity, obscuring which choices matter. We view supervised policies as distilling a decoding procedure through a teacher and examine independently what the teacher labels as committable, on which states the policy imitates it, what the policy observes, and how much the number of function evaluations (NFE) must decrease to offset policy overhead. We find that labeling predictions matching the final output is overly permissive at low NFE, and hidden states are weak alone but can complement confidence. About a 10% NFE reduction suffices to offset policy overhead for policies up to 7.5M parameters across tested settings. We therefore train a policy on trajectory-contraction targets, which accept only commitments staying on a reference trajectory, using confidence and hidden states. On LLaDA-8B-Instruct, the resulting 856K-parameter policy achieves the best accuracy--NFE frontier and highest throughput on GSM8K, is competitive at moderate targets on MATH-500 and HumanEval, and largely transfers to Dream-7B-Instruct.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Computer science, Offset (computer science), Complement (music), Matching (statistics), Language model, Decoding methods

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange