arXiv cs.AI· Ye Li, Jiahe Feng, Yuan Meng, Kangye Ji, Chen Tang, Jiajun Fan, Xinwan Wen, Jiawei Li, Wenwu Zhu, Shu-Tao Xia, Xin Wang, Zhi Wang·· 2 天前AI 评分34
TS-DP：面向时序自适应扩散策略加速的强化投机解码框架
TS-DP: Reinforcement Speculative Decoding For Temporal Adaptive Diffusion Policy Acceleration
AI 导读
研究团队提出 TS-DP，首个将投机解码用于 Diffusion Policy 的框架，通过蒸馏 Transformer drafter 替代高成本去噪调用，并用 RL 调度器按时序任务难度动态调整投机参数。仿真中实现最高 4.17x NFE 降低、3.2x 实际耗时加速，草稿接受率超 85%；真实实验 NFE 降低 3.8x，传送带任务成功率提升 23%，推理频率达 26 Hz。代码已开源。
正文 · 原文
Authors:Ye Li, Jiahe Feng, Yuan Meng, Kangye Ji, Chen Tang, Jiajun Fan, Xinwan Wen, Jiawei Li, Wenwu Zhu, Shu-Tao Xia, Xin Wang, Zhi Wang
Abstract:Diffusion Policy (DP) excels in embodied control but suffers from high inference latency and computational cost due to multiple iterative denoising steps. Existing diffusion model (DMs) acceleration methods improve inference speed by reducing computational overhead, as DMs for image generation are computation-bound. In this work, we reveal that DP are I/O bound, and thus propose using speculative decoding, a parallel computing framework, to accelerate DP for the first time. However, it is non-trivial to address the following challenges: (1) How to preserve the denoising quality of the base model at reduced computational cost under time-varying task difficulty in embodied settings; and (2) How to dynamically allocate computation according to task difficulty in such environments. Therefore, we propose Temporal-aware Reinforcement-based Speculative Diffusion Policy (TS-DP), the first framework that enables speculative decoding for DP with temporal adaptivity. First, to handle dynamic environments where task difficulty varies over time, we distill a Transformer-based drafter to imitate the base model and replace its costly denoising calls. Second, an RL-based scheduler further adapts to the time-varying task difficulty by adjusting speculative parameters to maintain accuracy while improving efficiency. Extensive experiments show that TS-DP achieves up to a 4.17x NFE reduction and a 3.2x wall-clock speedup with a draft acceptance rate of over 85% in simulation. Real-world experiments further demonstrate a 3.8x NFE reduction while improving success rates on real-time tasks (e.g., a 23% gain on the conveyor-belt task), reaching an inference frequency of 26 Hz and enabling real-time control while maintaining comparable task performance. The code is available at this https URL.
| Subjects: | Machine Learning (cs.LG); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2512.15773 [cs.LG] |
| (or arXiv:2512.15773v2 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2512.15773 arXiv-issued DOI via DataCite |
Submission history
From: Ye Li [view email]
[v1]
Sat, 13 Dec 2025 07:53:14 UTC (8,742 KB)
[v2]
Tue, 27 Oct 2026 02:37:43 UTC (28,808 KB)
来源：arXiv cs.AI · arxiv.org