arXiv Computation and Language

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

Informed Masking (IM) is a new technique for aligning Diffusion Large Language Models (dLLMs) with Reinforcement Learning (RL). It identifies a systematic upstream/downstream token structure in dLLM rollouts and shows that masking downstream tokens creates better subproblems for likelihood estimation. When integrated into three state‑of‑the‑art dLLM RL methods on LLaDA‑8B‑Instruct, IM yields up to 2.01%, 8.68%, and 5.77% relative average gains on math and planning benchmarks while improving training stability.

arXiv Machine Learning
Nov 6

PBPO: Prefix-Block Policy Optimization for Diffusion Language Models

The paper introduces Prefix-Block Policy Optimization (PBPO), a method that adapts block-level policy optimization to Diffusion Large Language Models (dLLMs). PBPO constructs a prefix-block state and uses a single model evaluation to create a factorized one‑step surrogate for block likelihood, allowing surrogate importance weighting and clipping to be applied separately to each token. Experiments on LLaDA-family models across six benchmarks and three generation lengths show that PBPO achieves competitive performance compared to existing dLLM reinforcement learning methods.

By Katherine Tieu, Dongqi Fu, Si Zhang, Hanqing Zeng, Tianxin Wei, Chenglin Wei, Yinglong Xia, Hong Li, Jingrui He
arXiv AI
Jun 2

SimSD: Simple Speculative Decoding in Diffusion Language Models

arXiv:2606. 02544v1 Announce Type: cross Abstract: Diffusion large language models (dLLMs) have recently emerged as a promising alternative to autoregressive (AR) LLMs, offering faster inference through parallel or blockwise decoding.

By Junxia Cui, Haotian Ye, Runchu Tian, Hongcan Guo, Jinya Jiang, Haoru Li, Chaojie Ren, Yiming Huang, Kaijie Zhu, Zhongkai Yu, Kun Zhou, Jingbo Shang
arXiv AI
Oct 16

TIPS: A Tiny Inference-time Policy for Scaling in Diffusion Language Models

The paper introduces TIPS, a tiny inference-time policy that controls which tokens are visible during each denoising pass of masked diffusion language models. TIPS, a 600‑K‑parameter shallow transformer, learns to unmask and remask tokens while keeping the language model frozen, using visibility‑aware confidence features and a fixed schedule for expected visibility. Trained via supervised warmup and reinforcement learning, TIPS achieves over 58% accuracy on GSM8K and MATH‑500 with only 32 decoder evaluations, surpassing strong baselines by more than 11 percentage points and remaining competitive at higher budgets without retraining.

By Ya-Shi Zhang, Jian Tang, Pierre-Andr\'e No\"el, Oleksiy Ostapenko
arXiv Machine Learning
Oct 8

CoDR: Training-Free Confidence-Drift Remasking for Diffusion Language Models

CoDR (Confidence Drift Remasking) is a training‑free, sampler‑agnostic refinement pass for masked diffusion language models that detects and corrects early token commitments that become unsupported as context grows denser. It estimates confidence drift for all committed positions using only a few forward passes and remasks only the tokens whose confidence has dropped, thereby improving accuracy across multiple backbones, tasks, and samplers with modest computational overhead.

By Yue Wu, Qinghe Zhang, Yu Zhang, Jian Huang
arXiv Computation and Language
Oct 22

Hierarchical Masked Diffusion Language Models

arXiv:2610.29255v1 Announce Type: new Abstract: Masked diffusion language models have emerged as an alternative to autoregressive models, enabling parallel token generation. However, standard masked...

By Parul Gupta, Satwik, Avni Aggarwal, Bhavesh Gurnani, Arnab Kumar Mondal, Dinesh Khandelwal, Dinesh Garg, Parag Singla