外观
论文arXiv 2505.219062025-05-28
ChatVLA-2发布：MoE架构保留VLM开放世界推理与数学能力
本文由 AI 根据公开内容整理，可能有误，请以原文为准：ChatVLA-2: Vision-Language-Action Model with Open-World Embodied Reasoning from Pretrained Knowledge（arXiv 2505.21906）
核心内容
针对现有端到端VLA模型在微调适配机器人任务时丢失预训练VLM核心能力的行业痛点，研究团队发布ChatVLA-2。该模型采用混合专家（MoE）架构并配套专用两阶段训练管线，旨在解决通用大模型向具身智能迁移过程中的“灾难性遗忘”问题。核心目标是确保VLA继承VLM的开放世界实体推理、数学求解及视觉空间智能，并能将其有效转化为机器人可执行动作序列，推动通用机器人基础模型发展。
技术验证设计了白板数学题匹配任务，要求机器人解读方程并抓取对应数字卡片。结果显示，ChatVLA-2在未进行针对性数学与OCR训练的前提下，展现出卓越的算术推理与文字识别能力，同时能理解涉及未见物体的新型方向指令。在综合推理与理解基准测试中，该方法显著超越OpenVLA、DexVLA和pi-zero等当前主流模仿学习SOTA方案，证实了MoE架构结合特定训练策略在保留大模型泛化能力方面的有效性。
关键看点
- MoE架构：采用混合专家模型耦合两阶段训练管线，解决VLA微调中VLM通用能力丢失问题
- 零样本数学推理：未显式训练即具备白板方程解读与数字卡片抓取能力，验证知识迁移效果
- 空间泛化能力：支持对训练集中未出现过的物体执行新型方向指令，突破 imitation learning 局限
- SOTA性能对比：在推理与理解任务上显著优于OpenVLA、DexVLA及pi-zero等主流模仿学习方法
工程与产业观察
该工作揭示了VLA研发从单纯追求动作模仿向保留认知推理能力转型的关键路径。通过MoE架构与分阶段训练解耦通用知识与具身技能，为构建兼具开放世界理解与精准操作能力的通用机器人大模型提供了可复现的工程范式，对解决长尾场景泛化难题具有重要参考价值。
英文摘要 (Abstract)
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.
作者：Zhongyi Zhou, Yichen Zhu, Junjie Wen, Chaomin Shen, Yi Xu
来源
本页事实对应的公开页面，以原文为准。