Think Less, Act Early: Reinforced Latent Reasoning with Early Exit in Vision-Language-Action Models
For embodied AI systems, this work addresses the computational inefficiency and error propagation of explicit reasoning, offering a faster and more stable alternative.
The paper proposes AVA-VLA, a latent reasoning framework for Vision-Language-Action models that replaces explicit Chain-of-Thought reasoning with reinforcement learning-optimized latent trajectories and an early-exit strategy, achieving 6x inference speedup and 98.3% success rate on LIBERO.
Existing Vision-Language-Action (VLA) models predominantly rely on explicit Chain-of-Thought (CoT) reasoning to bridge perception and action. While effective, this paradigm suffers from high computational costs and error propagation in multi-step tasks. In this paper, we propose Adaptive Variable Alignment VLA (AVA-VLA), a novel Latent Reasoning VLA framework that models reasoning as a sequence of unobservable latent variables, bypassing the need for explicit text generation. However, latent trajectories are inherently susceptible to noise interference and misalignment with downstream objectives. To address this, we introduce a Reinforcement Learning-based Denoising mechanism that treats latent state generation as a sequential decision process, optimizing reasoning trajectories via task-level rewards. Furthermore, we incorporate an Early-Exit Strategy that adaptively terminates reasoning based on state confidence, enabling a dynamic trade-off between depth and efficiency. Extensive experiments on embodied decision benchmarks demonstrate that AVA-VLA achieves a 6x inference speedup over explicit CoT methods while attaining a 98.3% average success rate on LIBERO, improving both efficiency and long-horizon stability over full-reasoning baselines.