arXiv cs.AI· Zixuan Liu, Nan Xue, Rui Bao, Yuhang Xu, Xiaofeng Cao, Jiangchao Yao, Zhiyong Chen·· 2 天前AI 评分27
LoopAhead：为 Looped Transformer 加速解码的推测框架
LoopAhead: Speculating Ahead for Faster Decoding in Looped Language Models
AI 导读
LoopAhead 是一个面向 Looped Transformer 的推测解码框架，通过轻量级前瞻草稿器（LAD）利用早期循环状态预测多个未来 token，并由深层循环反馈决定草稿分支的继续或暂停，最终由全深度模型做 token 决策。该方法在保持目标模型输出分布的同时，实现最高 7.82 倍的解码加速。
正文 · 原文
Abstract:Looped Transformers improve parameter efficiency by repeatedly applying the same Transformer layers. However, generating each token still involves sequential computation and repeated access to model weights, leading to high decoding latency. Recent pipelined self-speculative decoding methods reduce this delay by drafting tokens from early-loop predictions and asynchronously advancing tokens at different recurrent depths in the same batch, but they do not fully exploit early-loop features for longer lookahead and still spend computation verifying incorrect speculative branches. We introduce \method, a speculative decoding framework for looped transformers that enables earlier drafting of future tokens and assigns complementary roles to different recurrent depths. A lightweight lookahead drafter (LAD) predicts multiple future tokens with early-loop states, and deeper-loop feedback helps decide which draft branches to continue and which to pause, while the full-depth model makes the final token decisions. LoopAhead achieves up to 7.82x decoding speedup over autoregressive decoding while preserving the target model's output distribution.
| Subjects: | Artificial Intelligence (cs.AI); Machine Learning (cs.LG) |
| Cite as: | arXiv:2610.21594 [cs.AI] |
| (or arXiv:2610.21594v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2610.21594 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Zhiyong Chen [view email]
[v1]
Thu, 15 Oct 2026 09:38:11 UTC (602 KB)
来源：arXiv cs.AI · arxiv.org