arXiv AI

Fast-dLLM++: Fr\'echet Profile Decoding for Faster Diffusion LLM Inference

arXiv:2606. 02955v3 Announce Type: replace-cross Abstract: Diffusion large language models promise parallel token generation, yet inference remains bottlenecked by deciding which masked tokens can be safely committed together.

arXiv Computation and Language
Sep 21

Ripple-Pivot Search: Active Parallel Decoding for Diffusion Large Language Models

Ripple-Pivot Search (RPS) is a training‑free decoding method for Diffusion Large Language Models that identifies mid‑entropy pivot positions to reduce uncertainty across remaining masked tokens. By proactively committing these pivots and evaluating token assignments via lookahead, RPS enables more tokens to be unmasked in parallel, speeding up decoding. Experiments on three dLLMs and four reasoning/code‑generation benchmarks show 4–10× wall‑clock speedup over standard decoding, up to 18× with KV caching, while maintaining or improving generation quality.

By Yushi Ye, Xu Chen, Haoyun Jiang, Jinsong Lan, Haihong Tang, Xiangtao Li, Mingming Gong, Ivor Tsang, Yanfeng Wang, Jiangchao Yao
arXiv Computation and Language
Oct 21

TeCoP: Accelerating Diffusion Large Language Model Decoding via Tentative Context Pipelining

TeCoP is a training‑free decoder that accelerates diffusion large language model (dLLM) decoding by decoupling prediction readiness from context utility. It uses tentative context pipelining and draft‑and‑verify to reuse stable but uncommitted predictions as provisional cross‑block context, with commitment based on confidence and stability checks. Across four dLLMs on GSM8K, Minerva MATH, and MBPP, TeCoP cuts forward evaluations by up to 87% and boosts throughput up to 7.5‑fold without KV caching and 11.77‑fold with caching, outperforming Fast‑dLLM and DSB while maintaining accuracy within one percentage point.

By Jui-Chien Tsou, Hsuan-Chi Liu, Wei-Chin Wang, Yu-Syuan Xu, Chun-Yi Lee
arXiv AI
Sep 30

Reliable Parallel Decoding in Masked Diffusion Language Models

The paper introduces Reliable Parallel Decoding (RPD) for masked diffusion language models, addressing the unreliability of committing multiple predictions from a single forward pass. Diagnostics reveal that confidence alone is insufficient, as confident end‑sequence predictions can preempt necessary upstream computations, and downstream predictions degrade with upstream uncertainty. RPD selects candidates based on layer‑wise stability and final confidence, committing them under an entropy budget while deferring uncertain predictions, achieving superior throughput and competitive accuracy on LLaDA and Dream benchmarks.

By Zhenghao He, Bohan Liu, Guangzhi Xiong, Aidong Zhang
arXiv Computation and Language
Sep 23

PACE-dLLM: Elastic Block Decoding via Confidence Cliff Estimation for Diffusion Language Models

The paper introduces PACE-dLLM, an acceleration method for diffusion language models (dLLMs) that uses the model’s own per‑step confidence to estimate a ‘confidence cliff’ and determine the optimal look‑ahead horizon for block decoding. By fitting this cliff in closed form at each step, PACE-dLLM sets the horizon to its saturation point and applies an independent confidence threshold for token commitment, thereby avoiding the trade‑offs inherent in fixed‑size block decoding. Experiments on reasoning and code benchmarks show that PACE-dLLM achieves the best average accuracy on open‑source dLLM backbones while delivering significant wall‑clock speedups—up to 5.23× on LLaDA and 3.06× on Dream—improving the quality‑throughput Pareto frontier.

By Xiaocheng Lu, Shuhan Guo, Ziyue Ma, Jie Zhang, Jian Liu, Jingcai Guo, Haoxuan Che, Song Guo
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