arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2511.16166cs.CV
收藏

EvoVLA:自进化视觉-语言-动作模型

EvoVLA: Self-Evolving Vision-Language-Action Model

Peking University(北京大学)

Zeting Liu, Zida Yang, Zeyu Zhang, Hao Tang

AI总结:

EvoVLA通过自监督框架解决VLA模型的阶段幻觉问题,提升任务成功率和样本效率,实现有效的仿真到现实转移。

摘要:

长周期机械操作仍然是视觉-语言-动作(VLA)模型的挑战,尽管在零样本泛化和仿真到现实世界转移方面取得了最近的进步。当前的VLA模型存在阶段幻觉问题,即智能体利用粗略的评估信号来捷径多步骤任务,报告高进度但并未真正完成。我们提出了EvoVLA,一种自监督的VLA框架,通过三个互补的组件解决这一问题:阶段对齐奖励(SAR),利用Gemini生成的硬负样本进行三元组对比学习以防止视觉捷径;基于姿态的对象探索(POE),将好奇心根植于相对对象-夹具姿态而非原始像素;以及长周期记忆,利用选择性上下文保留和门控融合来稳定内在塑造过程。在Discoverse-L上进行的广泛评估显示,EvoVLA在平均任务成功率方面比最强基线(OpenVLA-OFT)提高了10.2个百分点,达到69.2%。EvoVLA还实现了1.5倍更好的样本效率,并将阶段幻觉从38.5%降至14.8%。在物理机器人上的现实世界部署在四个操作任务上的平均成功率为54.6%,优于OpenVLA-OFT 11分,证明了有效的仿真到现实转移和强大的泛化能力。代码:https://github.com/AIGeeksGroup/EvoVLA。网站:https://aigeeksgroup.github.io/EvoVLA。

英文摘要:

Long-horizon robotic manipulation remains challenging for Vision-Language-Action (VLA) models despite recent progress in zero-shot generalization and simulation-to-real-world transfer. Current VLA models suffer from stage hallucination, where agents exploit coarse evaluation signals to shortcut multi-step tasks, reporting high progress without truly completing them. We present EvoVLA, a self-supervised VLA framework that addresses this issue through three complementary components: Stage-Aligned Reward (SAR), which uses triplet contrastive learning with Gemini-generated hard negatives to prevent visual shortcuts; Pose-Based Object Exploration (POE), which grounds curiosity in relative object-gripper pose instead of raw pixels; and Long-Horizon Memory, which uses selective context retention and gated fusion to stabilize intrinsic shaping during extended rollouts. Extensive evaluations on Discoverse-L, a long-horizon manipulation benchmark with three multi-stage tasks, show that EvoVLA improves average task success by 10.2 percentage points over the strongest baseline (OpenVLA-OFT), reaching 69.2 percent. EvoVLA also achieves one-and-a-half times better sample efficiency and reduces stage hallucination from 38.5 percent to 14.8 percent. Real-world deployment on physical robots reaches an average success rate of 54.6 percent across four manipulation tasks, outperforming OpenVLA-OFT by 11 points, demonstrating effective sim-to-real transfer and strong generalization. Code: https://github.com/AIGeeksGroup/EvoVLA. Website: https://aigeeksgroup.github.io/EvoVLA.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑