arXiv cs.AI· Yixue Zhang, Kun Wu, Zhi Gao, Zhen Zhao, Pei Ren, Zhiyuan Xu, Fei Liao, Xinhua Wang, Shichao Fan, Di Wu, Qiuxuan Feng, Meng Li, Zhengping Che, Chang Liu, Jian Tang·· 4 小时前AI 评分48
RoboGene：面向 VLA 预训练的多样性驱动智能体任务生成框架
RoboGene: Diversity-Driven Agentic Task Generation Framework for VLA Pre-training
AI 导读
RoboGene 是一个用于生成多样、有信息量机器人操作任务的智能体框架，覆盖单臂、双臂和移动机器人，结合多样性驱动采样、智能体自我反思的质量评估和人机校验反馈记忆。该工作收集了 18k 条真实世界轨迹，并提出任务可行性与多样性指标，表现显著优于 GPT-4o 和 Gemini 2.5 Pro。用其生成数据训练的 VLA 模型成功率更高、泛化更强，论文已被 CoRL 2026 接收。
正文 · 原文
Authors:Yixue Zhang, Kun Wu, Zhi Gao, Zhen Zhao, Pei Ren, Zhiyuan Xu, Fei Liao, Xinhua Wang, Shichao Fan, Di Wu, Qiuxuan Feng, Meng Li, Zhengping Che, Chang Liu, Jian Tang
Abstract:The pursuit of general-purpose robotic manipulation is hindered by the scarcity of diverse, real-world interaction data. Unlike large-scale web vision-language data, robotic data collection incurs prohibitive physical costs. The key challenge is thus not merely to collect more robot data, but to obtain informative training data that broadens behavioral coverage across objects, scenes, and skills. Yet automatically curating task proposals to maximize the utility of collected data remains under-explored.
We introduce RoboGene, an agentic framework for generating diverse, informative manipulation tasks across single-arm, dual-arm, and mobile robots. RoboGene combines (1) diversity-driven sampling for broad task coverage, (2) agentic self-reflection for task-quality assessment, and (3) human-verified feedback memory for execution-grounded validation and continual refinement.
Through quantitative analysis and large-scale real-world experiments, we collect 18k trajectories and introduce metrics for task feasibility and diversity. RoboGene significantly outperforms frontier foundation models, including GPT-4o and Gemini 2.5 Pro. VLA models trained on RoboGene-generated data achieve higher success rates and stronger generalization, highlighting the value of high-quality task generation.
Project page: this https URL. Datasets: Hugging Face (this https URL) and ModelScope (this https URL).
| Comments: | Accepted at the 10th Conference on Robot Learning (CoRL 2026). 36 pages |
| Subjects: | Robotics (cs.RO); Artificial Intelligence (cs.AI); Machine Learning (cs.LG) |
| Cite as: | arXiv:2602.16444 [cs.RO] |
| (or arXiv:2602.16444v3 [cs.RO] for this version) | |
| https://doi.org/10.48550/arXiv.2602.16444 arXiv-issued DOI via DataCite |
Submission history
From: Yixue Zhang [view email]
[v1]
Wed, 18 Feb 2026 13:29:43 UTC (26,299 KB)
[v2]
Thu, 19 Feb 2026 04:26:20 UTC (26,299 KB)
[v3]
Mon, 26 Oct 2026 06:27:54 UTC (20,245 KB)
来源：arXiv cs.AI · arxiv.org