MindExplore+: Towards Long-Horizon Vision-Language-Action System

Vision-Language-Action (VLA) is crucial for autonomous decision-making in embodied systems. While current methods have advanced single-skill abilities, their short-horizon capability limits applicability in real-world scenarios. To address this challenge, we propose MindExplore+, a hierarchical VLA system with cross-skill capabilities for long-horizon tasks in dynamic environments. By iteratively aligning task planning and action execution, this approach ensures excellent generalization across diverse real-world scenarios. In the reasoning layer, task specific chains of thought (CoT) are designed for planning long horizon task sequences and providing meta-action signals. In the acting layer, a simple but powerful Mixture of Policy Experts strategy is built inspired by signals and multimodal inputs for adaptively selecting skill experts and generating closed-loop action sequences. Also, it presents a lightweight Multimodal Diffusion Policy (MMDP) to enhance spatial perception by fusing multi-visual modality features. Besides, the pioneering memory mechanism establishes feedback between the reasoning and acting layers, facilitating adaptive execution of long-horizon tasks and real-time replanning. Notably, we create SandGo-1k and SandThink-21k, the first expert-level multimodal embodied dataset and CoT dataset tailored for sandy environments. To further improve efficiency in embodied tasks, we introduce a Biview Token Optimization (BTO) strategy. The method prunes redundant tokens based on a set of pivot tokens to ensure preserving diversity, then low-attention tokens are aggregated to ensure preserving task relevance while further reducing sequence length. MindExplore+ is 3.01× more successful than existing methods in unstructured and dynamic environments. Furthermore, by using BTO, it achieves a 1.64× speedup and reduces FLOPs to 18.85% of the original model, while maintaining competitive task performance.