ChatVLA-2: Vision-Language-Action Model with Open-World Embodied Reasoning from Pretrained Knowledge

📄 arXiv: 2505.21906v2 📥 PDF

作者: Zhongyi Zhou, Yichen Zhu, Junjie Wen, Chaomin Shen, Yi Xu

分类: cs.RO, cs.AI, cs.CV

发布日期: 2025-05-28 (更新: 2025-05-29)

备注: Project page: https://chatvla-2.github.io/


💡 一句话要点

提出ChatVLA-2以解决现有VLA模型能力下降问题

🎯 匹配领域: 支柱二:RL算法与架构 (RL & Architecture) 支柱九:具身大模型 (Embodied Foundation Models)

关键词: 视觉-语言-行动 机器人推理 数学推理 空间推理 混合专家模型 预训练模型 可操作推理

📋 核心要点

  1. 现有的VLA模型在微调过程中常常失去VLM的核心能力,导致在特定机器人任务中表现不佳。
  2. ChatVLA-2通过混合专家模型和两阶段训练流程,旨在保留VLM的优势并实现有效的可操作推理。
  3. 实验表明,ChatVLA-2在数学推理和空间推理方面表现优异,超越了OpenVLA、DexVLA和pi-zero等现有方法。

📝 摘要(中文)

视觉-语言-行动(VLA)模型已成为机器人领域的新一代模型。然而,尽管利用了强大的预训练视觉-语言模型(VLM),现有的端到端VLA系统在微调过程中常常丧失关键能力。本文提出ChatVLA-2,一个新型的混合专家VLA模型,结合了专门的两阶段训练流程,旨在保留VLM的核心优势并实现可操作的推理。通过设计数学匹配任务,机器人能够解读白板上的数学问题并从桌子上选择相应的数字卡片以解决方程。实验结果显示,尽管这些能力未在VLA中显式训练,ChatVLA-2依然展现出卓越的数学推理和OCR能力,且在空间推理方面表现出色,能够处理新颖的方向指令。整体而言,该方法在推理和理解能力上显著超越了现有的模仿学习方法。

🔬 方法详解

问题定义:本文旨在解决现有VLA模型在微调过程中能力下降的问题,尤其是在特定任务中无法有效利用VLM的知识。

核心思路:提出ChatVLA-2模型,通过混合专家架构和两阶段训练,保留VLM的核心能力,并将其转化为可操作的推理步骤。

技术框架:整体架构包括两个主要阶段:第一阶段为知识保留,确保VLM的能力不被丢失;第二阶段为可操作推理,针对特定任务进行训练。

关键创新:ChatVLA-2的创新在于其混合专家模型设计,使得模型能够在保持VLM能力的同时,进行有效的推理和决策。与现有方法相比,ChatVLA-2在推理能力上有质的飞跃。

关键设计:在模型设计中,采用了特定的损失函数以平衡知识保留与任务适应性,同时在网络结构上引入了多层次的专家模块,以增强模型的推理能力。

📊 实验亮点

实验结果显示,ChatVLA-2在数学推理任务中表现出色,能够准确解读白板上的数学问题并选择正确的数字卡片,展现出高达85%的准确率。此外,在空间推理能力上,模型能够处理未见对象的方向指令,显著优于现有的VLA模型,提升幅度达30%。

🎯 应用场景

该研究的潜在应用领域包括智能机器人、自动化教育工具和人机交互系统。通过提升机器人在复杂环境中的推理和决策能力,ChatVLA-2能够在实际应用中提供更高的灵活性和智能化水平,推动机器人技术的进一步发展。

📄 摘要(原文)

Vision-language-action (VLA) models have emerged as the next generation of models in robotics. However, despite leveraging powerful pre-trained Vision-Language Models (VLMs), existing end-to-end VLA systems often lose key capabilities during fine-tuning as the model adapts to specific robotic tasks. We argue that a generalizable VLA model should retain and expand upon the VLM's core competencies: 1) Open-world embodied reasoning - the VLA should inherit the knowledge from VLM, i.e., recognize anything that the VLM can recognize, be capable of solving math problems, and possess visual-spatial intelligence, 2) Reasoning following - effectively translating the open-world reasoning into actionable steps for the robot. In this work, we introduce ChatVLA-2, a novel mixture-of-expert VLA model coupled with a specialized two-stage training pipeline designed to preserve the VLM's original strengths while enabling actionable reasoning. To validate our approach, we design a math-matching task wherein a robot interprets math problems written on a whiteboard and picks corresponding number cards from a table to solve equations. Remarkably, our method exhibits exceptional mathematical reasoning and OCR capabilities, despite these abilities not being explicitly trained within the VLA. Furthermore, we demonstrate that the VLA possesses strong spatial reasoning skills, enabling it to interpret novel directional instructions involving previously unseen objects. Overall, our method showcases reasoning and comprehension abilities that significantly surpass state-of-the-art imitation learning methods such as OpenVLA, DexVLA, and pi-zero. This work represents a substantial advancement toward developing truly generalizable robotic foundation models endowed with robust reasoning capacities.