arXiv cs.AI· Chuanke Pang, Junyi Huang, Jianye Fang, Zhijun Zhao, Yaobing Wang, Kun Xu, Xilun Ding·· 11 小时前AI 评分40
InDex：通过意图条件微调让 VLA 模型适配灵巧操作
Bridging the Morphology Gap: Adapting VLA Models to Dexterous Manipulation via Intent-Conditioned Fine-Tuning
AI 导读
研究提出 InDex 框架，用归一化开合信号监督标量抓取意图，分两阶段将并行夹爪 VLA 模型适配到多指灵巧手。在四个仿真任务中，InDex 搭配 π₀.₅ 取得 85.8% 平均任务成功率，比原生策略提升 35.5 个百分点，并在未微调的相似几何物体上达到 69.0%。该方法在三个 VLA 主干和真实世界策略上同样有效，平均成功率从 71.0% 升至 83.0%。
正文 · 原文
Abstract:Pretrained vision-language-action (VLA) models offer semantic and spatial priors for robot manipulation, but their parallel-gripper interfaces leave multi-finger contact coordination implicit. Adapting these models requires both timely grasp transitions and embodiment-specific hand trajectories. We introduce InDex, an intent-conditioned framework that connects these two requirements. A normalized aperture signal, derived from retargeted demonstrations, supervises a scalar grasp intent while preserving full hand-action supervision. Stage 1 learns synchronized end-effector-intent chunks; Stage 2 conditions a diffusion hand decoder on these predictions, visual-language context, and proprioception. Sequential optimization establishes the upstream interface before learning dexterous control with low-rate co-adaptation. Across four simulated tasks, InDex with $\pi_{0.5}$ achieves 85.8% mean task success, improving the native policy by 35.5 percentage points, and reaches 69.0% on held-out objects with similar geometry without fine-tuning or recalibration. Improvements also hold across three VLA backbones and separately trained real-world policies, where mean success increases from 71.0% to 83.0%. Component ablations and timing interventions support the value of explicit grasp-intent conditioning and expressive hand generation. These results identify arm-hand coordination as an actionable bottleneck in adapting parallel-gripper priors to dexterous manipulation.
| Subjects: | Robotics (cs.RO); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2606.12109 [cs.RO] |
| (or arXiv:2606.12109v3 [cs.RO] for this version) | |
| https://doi.org/10.48550/arXiv.2606.12109 arXiv-issued DOI via DataCite |
Submission history
From: Chuanke Pang [view email]
[v1]
Wed, 10 Jun 2026 14:03:52 UTC (14,276 KB)
[v2]
Tue, 28 Jul 2026 15:33:03 UTC (18,657 KB)
[v3]
Wed, 4 Nov 2026 07:20:53 UTC (26,552 KB)
来源：arXiv cs.AI · arxiv.org