arXiv cs.AI· Yixiang Zhu, Yonghao Chen, Jiaxiang Zou, Yusong Hu, Xinyu Chen·· 13 小时前AI 评分34
DEFLECT：面向延迟鲁棒异步 VLA 的时间反事实偏好学习
DEFLECT: Temporal Counterfactual Preference Learning for Delay-Robust Asynchronous VLAs
AI 导读
DEFLECT 是一个离线后训练框架，利用训练轨迹中可观测的预测-执行状态错配构建偏好监督，提升异步 VLA 策略的延迟鲁棒性。在 Kinetix、LIBERO 和三个真实机器人任务上，最长支持延迟下 Kinetix 成功率比 VLASH 高 4.2 个百分点，LIBERO 上比 GR00T N1.7 的朴素异步执行高 22.7 个百分点。
正文 · 原文
Abstract:Vision-Language-Action (VLA) policies increasingly rely on asynchronous inference to hide model latency behind ongoing robot motion. During inference, the robot continues executing its current action chunk while the next chunk is generated from the observation available at inference start. By the time the new chunk is executed, the robot or scene may have changed, so an action that fits the prediction-time state can be misaligned with the execution-time state. Training-time methods modify the policy input or supervised target based on the expected execution delay, but do not directly supervise how actions should change as the scene evolves between prediction and execution. Runtime repair methods improve chunk continuity or modify actions during deployment, but can still leave actions aligned with the prediction-time state rather than the execution-time state. Our key insight is that execution-time observations are unavailable when the online prediction is made but are available in recorded trajectories during offline training, making the prediction-execution mismatch observable during training. We propose DEFLECT, an offline post-training framework that turns this temporal contrast into preference supervision for a policy acting from stale inputs. Across Kinetix, LIBERO, and three real-robot tasks, DEFLECT improves delay robustness across the evaluated VLA policies and settings. At the longest supported delay, it improves success by 4.2 percentage points over VLASH on Kinetix and by 22.7 points over naive asynchronous execution with GR00T N1.7 on LIBERO.
| Subjects: | Robotics (cs.RO); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2605.19294 [cs.RO] |
| (or arXiv:2605.19294v3 [cs.RO] for this version) | |
| https://doi.org/10.48550/arXiv.2605.19294 arXiv-issued DOI via DataCite |
Submission history
From: Yixiang Zhu [view email]
[v1]
Tue, 19 May 2026 03:14:11 UTC (561 KB)
[v2]
Wed, 3 Jun 2026 03:32:00 UTC (13,769 KB)
[v3]
Tue, 20 Oct 2026 10:44:37 UTC (9,073 KB)
来源：arXiv cs.AI · arxiv.org