paper-with-me

Papers

d3LLM: Ultra-Fast Diffusion LLM using Pseudo-Trajectory Distillation

2026-01-12 · Yu-Yang Qian, Junda Su, Lanxiang Hu, Peiyuan Zhang, Zhijie Deng, Peng Zhao, Hao Zhang arxiv

Diffusion large language models (dLLMs) offer capabilities beyond those of autoregressive (AR) LLMs, such as parallel decoding and random-order generation. However, realizing these benefits in practice is non-trivial, as dLLMs inherently face an accuracy-parallelism trade-off. Despite increasing interest, existing methods typically focus on only one-side of the coin, targeting either efficiency or accuracy. To address this limitation, we propose d3LLM (Pseudo-Distilled Diffusion Large Language Model), striking a balance between accuracy and parallelism: (i) during training, we introduce pseudo-trajectory distillation to teach the model which tokens can be decoded confidently at early steps, thereby improving parallelism; (ii) during inference, we employ entropy-based multi-block decoding with a KV-cache refresh mechanism to achieve high parallelism while maintaining accuracy. To better evaluate dLLMs, we also introduce AUP (Accuracy Under Parallelism), a new metric that jointly measures accuracy and parallelism. Experiments demonstrate that our d3LLM achieves up to 10$\times$ speedup over vanilla LLaDA/Dream, and 5$\times$ speedup over AR models without much accuracy drop. Our code is available at https://github.com/hao-ai-lab/d3LLM.

📄 PDF Abstract BibTeX arXiv:2601.07568

Code (0)

등록된 구현이 없습니다.

Similar Papers 제목 키워드 기반

dUltra: Ultra-Fast Diffusion Language Models via Reinforcement Learning

2025-12-24 · Shirui Chen, Jiantao Jiao, Lillian J. Ratliff, Banghua Zhu arxiv

Masked diffusion language models (MDLMs) offer the potential for parallel token generation, but most open-source MDLMs decode fewer than 5 tokens per model forward pass even with sophisticated sampling strategies, limiti…

Reinforcement LearningMathematical ReasoningCode Generation

Streaming Diffusion Policy: Fast Policy Synthesis with Variable Noise Diffusion Models

2024-06-07 · Sigmund H. Høeg, Yilun Du, Olav Egeland

Diffusion models have seen rapid adoption in robotic imitation learning, enabling autonomous execution of complex dexterous tasks. However, action synthesis is often slow, requiring many steps of iterative denoising, lim…

DenoisingImage GenerationImitation LearningRobot Manipulation

Distribution Backtracking Builds A Faster Convergence Trajectory for Diffusion Distillation

2024-08-28 · Shengyuan Zhang, Ling Yang, Zejian Li, An Zhao 외

Accelerating the sampling speed of diffusion models remains a significant challenge. Recent score distillation methods distill a heavy teacher model into a student generator to achieve one-step generation, which is optim…

Ultra Fast PDE Solving via Physics Guided Few-step Diffusion

2026-02-03 · Cindy Xiangrui Kong, Yueqi Wang, Haoyang Zheng, Weijian Luo 외 arxiv

Diffusion-based models have demonstrated impressive accuracy and generalization in solving partial differential equations (PDEs). However, they still face significant limitations, such as high sampling costs and insuffic…

TraFlow: Trajectory Distillation on Pre-Trained Rectified Flow

2025-02-24 · Zhangkai Wu, Xuhui Fan, Hongyu Wu, Longbing Cao

Majorities of distillation methods on pre-trained diffusion models or on pre-trained rectified flow, focus on either the distillation outputs or the trajectories between random noises and clean images to speed up sample …