Mantis: A Versatile Vision-Language-Action Model with Disentangled Visual Foresight

TL;DR

Mantis introduces a disentangled visual foresight framework that enhances vision-language-action models by improving comprehension, reasoning, and action prediction through a novel decoupled architecture and training on diverse datasets.

Contribution

The paper proposes a novel Disentangled Visual Foresight (DVF) framework with meta queries and a diffusion Transformer head, enabling better visual state prediction and reasoning in VLA models.

Findings

Achieves 96.7% success rate on LIBERO benchmark after fine-tuning.

Outperforms existing models in instruction-following and generalization.

Demonstrates high convergence speed and improved reasoning capabilities.

Abstract

Recent advances in Vision-Language-Action (VLA) models demonstrate that visual signals can effectively complement sparse action supervisions. However, letting VLA directly predict high-dimensional visual states can distribute model capacity and incur prohibitive training cost, while compressing visual states into more compact supervisory signals inevitably incurs information bottlenecks. Moreover, existing methods often suffer from poor comprehension and reasoning capabilities due to the neglect of language supervision. This paper introduces Mantis, a novel framework featuring a Disentangled Visual Foresight (DVF) to tackle these issues. Specifically, Mantis decouples visual foresight prediction from the backbone with the combination of meta queries and a diffusion Transformer (DiT) head. With the current visual state provided to the DiT via a residual connection, a simple next-state prediction objective enables the meta queries to automatically capture the latent actions that delineate the visual trajectory, and hence boost the learning of explicit actions. The disentanglement reduces the burden of the VLA backbone, enabling it to maintain comprehension and reasoning capabilities through language supervision. Empirically, pretrained on human manipulation videos, robot demonstrations, and image-text pairs, Mantis achieves a 96.7% success rate on LIBERO benchmark after fine-tuning, surpassing powerful baselines while exhibiting high convergence speed. Real-world evaluations show that Mantis outperforms $\pi_{0.5}$, a leading open-source VLA model, particularly in instruction-following capability, generalization to unseen instructions, and reasoning ability. Code and weights are released to support the open-source community.