InDex: Intent-Conditioned Fine-Tuning for Dexterous Manipulation
What happened
Chuanke Pang and other authors propose the InDex framework, which uses normalized open-close signals obtained from retargeted demonstrations to supervise a scalar grasp intent, adapting a parallel-jaw-gripper VLA model to a multi-fingered dexterous hand in two stages. Across four simulated tasks, InDex with π0.5 achieves an average task success rate of 85.8%, 35.5 percentage points higher than the native policy; on new objects that are geometrically similar but not fine-tuned, the success rate is 69.0%. The authors state that this improvement also holds across three VLA backbones and a separately trained real-world policy, whose average success rate rises from 71.0% to 83.0%.
AI 根据报道生成 · 6 days ago更新
Reporting timeline
Follow the reporting to explore different perspectives.
- arXiv cs.ROInDex: Adapting VLA models to dexterous manipulation via intent-conditioned fine-tuning
The study proposes the InDex framework, which uses a normalized aperture signal derived from retargeted demonstrations to supervise a scalar grasp intent, adapting parallel-gripper VLA models to multi-fingered dexterous hands in two stages. Across four simulated tasks, InDex with π0.5 achieved 85.8% mean task success, a 35.5 percentage point improvement over the native policy, and reached 69.0% on new objects with similar geometry without fine-tuning. It was also effective across three VLA backbones and on a separately trained real-world policy, whose mean success rate rose from 71.0% to 83.0%.
Activity for this event
- Current
- 3
- Peak
- 10Nov 5 14:00
- Past 24 hours
- -49%
The trend compares participants observed throughout the same period, so its range may be smaller than the current interest count. Move or tap to inspect each hour; use the arrow keys on a keyboard.