InternVLA-A1.5: Unifying Understanding, Latent Foresight, and Action for Compositional Generalization
2026-07-06 • Robotics
Robotics
AI summaryⓘ
The authors developed InternVLA-A1.5, a robot control model that combines language understanding from pretrained vision-language models (VLMs) with physical action prediction. Instead of learning future events pixel by pixel, their method uses a compact representation guided by a pretrained video model to capture future dynamics without slowing down the robot. This approach keeps the model's language understanding intact and helps it perform well on diverse simulation tasks and real-world instructions. Their design allows for smoother long-term planning and better generalization to new instructions.
Vision-Language ModelsRobot ManipulationFuture PredictionVideo GenerationLatent RepresentationMultimodal LearningCompositional GeneralizationReal-Time ControlWorld-Model DynamicsSubtask Prediction
Authors
Haoxiang Ma, Junhao Cai, Xiaoxu Xu, Hao Li, Yuyin Yang, Yang Tian, Jiafei Cao, Hongrui Zhu, Zherui Qiu, Zhaxizhuoma, Yuqiang Yang, Jiaqi Peng, Xueyuan Wei, Yangkun Zhu, Jiahao Jiang, Xing Gao, Hanqing Wang, Feng Yuan, Kailin Li, Xueyue Zhu, Tai Wang, Yan Ding, Jiangmiao Pang, Jia Zeng, Jingjing Zhang, Bowen Zhou, Yao Mu, Chunhua Shen, Weinan Zhang
Abstract
Unified models for robot manipulation aim to equip one policy with both the semantic priors of pretrained VLMs and the physical dynamics learned through future prediction. In practice, existing designs tend to erode the semantics of the pretrained backbone, suffer interference among heterogeneous objectives, and learn future prediction from scratch in pixel space, leaving the dynamics priors of pretrained video generators unexploited. We present InternVLA-A1.5, which builds the policy on a native VLM backbone that keeps training on VQA and subtask prediction, and attaches a lightweight unified expert for continuous action generation. Future prediction is recast as a latent-querying problem, where a small set of learnable foresight tokens condenses the task-relevant future into a compact latent code under the supervision of a frozen pretrained video generation model, so the policy inherits world-model dynamics priors without ever learning pixel-level generation. The video branch is discarded at inference, keeping real-time control. Pretrained on 1.2M robot episodes and 3M multimodal samples, InternVLA-A1.5 achieves the best overall results on all six simulation benchmarks. In the real world, the preserved semantics deliver the strongest compositional generalization on held-out instruction bindings, and the two designs together sustain long-horizon execution.