BiDex-100 benchmarks bimanual dexterous manipulation across 106 simulation tasks
Original title: BiDex-100: A Comprehensive Benchmark for Bimanual Dexterous Manipulation Across 100+ Simulation Tasks
Each task has an out-of-domain variant, and seven vision-language-action models are scored on 8,080 demonstrations.
BiDex-100 is a simulation benchmark for two-handed dexterous manipulation with high-degree-of-freedom hands. It includes 106 tasks on five axes: generalization, instruction following, limited demonstrations, long horizon, and memory. Each task has an out-of-domain variant with controlled visual and spatial changes. The release provides two embodiments and 8,080 success-validated teleoperated demonstrations totaling 6.25 million frames, plus baselines for seven recent vision-language-action models.
- Tasks cover generalization, instruction following, few demonstrations, long horizon, and memory.
- Demonstrations total 6.25 million frames across two hand embodiments.
- Seven vision-language-action models are scored for robustness and data efficiency.
From the original article
1 min read at huggingface.co
Bimanual manipulation with dexterous hands has attracted growing attention in research on robotic action models. However, few studies have developed benchmarks specifically tailored for dexterous manipulation, leaving the dexterous manipulation capabilities of existing models insufficiently characterized.
By Yong Wang, Kecheng Zheng, Xing Zhu · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.