Skip to content

论文arXiv 2611.148349 天前

HDH框架利用VLA历史数据训练轻量验证器提升操作成功率 ​

本文由 AI 根据公开内容整理,可能有误,请以原文为准:Learning Low-Latency Action Verification for Frozen Vision-Language-Action Models(arXiv 2611.14834)

核心内容 ​

针对视觉-语言-动作(VLA)模型在通用机器人操作中存在的动作序列无效及验证器计算昂贵导致的高延迟问题,研究团队提出HDH轻量级框架。该方案核心在于不微调VLA基座,而是利用其自身部署产生的回溯数据(Hindsight)来优化动作选择。实验显示,HDH在LIBERO-10仿真基准中实现82.3%的平均成功率,并在三项真实世界操作任务中使平均成功率较基线策略显著提升22.6%。

HDH技术架构包含三阶段:首先部署冻结的VLA策略并收集自主运行轨迹;其次训练一个紧凑的动作验证器,为每个动作块预测任务成功分数;最后在推理时由VLA生成多个候选动作,经该轻量验证器评分后择优执行。这种“生成-验证”解耦机制避免了在每个策略查询节点引入重型计算,有效解决了VLA落地中的实时性与可靠性矛盾,为具身智能端侧部署提供了低成本纠错范式。

关键看点 ​

  • 回溯数据利用:基于冻结VLA自身部署轨迹训练验证器,无需额外标注或基座微调
  • 推理加速机制:采用紧凑验证器对动作块打分择优,替代高算力消耗的传统评估流程
  • 真机性能增益:在三项真实操作任务中平均成功率较基线提升22.6%,验证工程有效性
  • 仿真基准表现:LIBERO-10任务平均成功率达82.3%,确立轻量化验证方案的技术上限

工程与产业观察 ​

HDH证明了在不增加VLA参数量前提下,通过挖掘模型自身历史交互数据训练专用验证器,可显著弥补端到端策略的动作采样缺陷。该思路为具身大模型从仿真走向真机提供了关键的“安全兜底”与“质量过滤”工程路径,降低了高精度操作对超大算力的依赖,对VLA模型的边缘侧实时部署具有重要参考价值。

英文摘要 (Abstract) ​

Vision-language-action (VLA) models have made significant progress in general-purpose robotic manipulation. However, sampled action sequences can be ineffective, while evaluating alternatives with computationally expensive verifiers adds latency at each policy query. In this work, we propose HDH, a lightweight framework that uses hindsight from a frozen VLA's own deployments to improve action verification. The framework has three stages. First, we deploy the policy and collect its autonomous rollouts. We then train a compact action verifier to assign each action chunk a predicted task-success score. Finally, the frozen VLA generates action candidates, and the verifier selects the candidate with the highest score for execution. Across four LIBERO-10 tasks, HDH achieves an average success rate of 82.3% and on three real-world manipulation tasks, HDH improves average success over the baseline policy by 22.6%.

作者:Wen-Han Hsieh, Ken Jiankun Zheng, Chia-Chien Chen, David M. Chan, Trevor Darrell

基于本页技术方向关联的整机参数、核心零部件、实测报告与同赛道选型横评。

外部来源

本页事实对应的公开页面,以原文为准。(点击将离开本站)

  1. arXiv 2611.14834:Learning Low-Latency Action Verification for Frozen Vision-Language-Action Models ↗访问 2026-11-14