paper-with-me

Papers

Multi-Token Residual Prediction

2026-05-12 · Yufeng Xu, Zishuo Bao, Qian Wang, Zeshen Zhang, Haoqi Zhang, Bowen Peng, Ang Li, Rahul Chalamala, Yucheng Lu arxiv

Diffusion Language Models (DLMs) generate text by iteratively denoising masked token sequences, offering a tradeoff between parallelism and quality compared to autoregressive models. In current practice, the number of tokens decoded per step is controlled by a confidence threshold, and quality degrades monotonically as more tokens are denoised per step. We introduce Multi-token Residual Prediction (MRP), a lightweight module that enables dependency-aware multi-token denoising within a single backbone forward pass. MRP exploits a key property of the denoising process: the logit distributions at adjacent denoising steps are remarkably similar. Rather than running the backbone a second time to obtain the next-step logits, MRP predicts the residual between steps from the backbone's hidden states, effectively denoising more tokens per backbone forward at a fraction of the cost. We apply MRP across the two operating regimes of DLM decoding. In the high-quality-low-throughput static denoising regime, MRP serves as a drafter for speculative decoding: its proposals are verified against the backbone, yielding lossless acceleration of up to 1.4x in SGLang. In the low-quality-high-throughput dynamic denoising regime, MRP instead drives a remasking scheme that revokes over-eager reveals, recovering most of the accuracy lost to aggressive low-threshold decoding and improving accuracy by up to 22.6 points on code generation task HumanEval and 17.7 points on reasoning task GSM8K.

📄 PDF Abstract BibTeX arXiv:2605.18817

Code (0)

등록된 구현이 없습니다.

Tasks

Code Generation

Similar Papers 제목 키워드 기반

Residual Connections and the Causal Shift: Uncovering a Structural Misalignment in Transformers

2026-02-16 · Jonathan Lys, Vincent Gripon, Bastien Pasdeloup, Axel Marmoret 외 arxiv

Large Language Models (LLMs) are trained with next-token prediction, implemented in autoregressive Transformers via causal masking for parallelism. This creates a subtle misalignment: residual connections tie activations…

Transformers represent belief state geometry in their residual stream

2024-05-24 · Adam S. Shai, Sarah E. Marzen, Lucas Teixeira, Alexander Gietelink Oldenziel 외

What computational structure are we building into large language models when we train them on next-token prediction? Here, we present evidence that this structure is given by the meta-dynamics of belief updating over hid…

Prediction

KunlunBaize: LLM with Multi-Scale Convolution and Multi-Token Prediction Under TransformerX Framework

2025-02-27 · Cheng Li, Jiexiong Liu, Yixuan Chen, Yanqin Jia 외

Large language models have demonstrated remarkable performance across various tasks, yet they face challenges such as low computational efficiency, gradient vanishing, and difficulties in capturing complex feature intera…

Computational Efficiency

HyperDFlash: Hyper-Connection-Aligned Block Speculative Decoding with Gated Residual Reduction

2026-06-25 · Luxi Lin, Shuang Peng, Rui Ma, Junhao Hua 외 arxiv

We present HyperDFlash, a block-parallel speculative decoding framework tailored to DeepSeek-V4's Hyper-Connections (HC). Despite the strong performance of DeepSeek-V4's native Multi-Token Prediction (MTP) module on init…

X-Tokenizer: A Multimodal Action Tokenizer for Vision-Language-Action Pretraining

2026-06-07 · Miracle Kang, Lights Shi, Lucy Liang, Roy Gan 외 arxiv

Modern Vision-Language-Action (VLA) models must bridge pretrained vision-language reasoning and precise continuous robot control. Existing action tokenizers discretize actions primarily for reconstruction, producing code…

Multimodal Reasoning