Fast-ThinkAct: Efficient Vision-Language-Action Reasoning via Verbalizable Latent Planning

2026-01-14

AI Summary

This work addresses the high inference latency and efficiency–performance trade-off inherent in existing Vision-Language-Action (VLA) methods that rely on explicit chains of thought. To overcome this limitation, we propose Fast-ThinkAct, a novel framework that introduces, for the first time, a verbalizable implicit chain of thought. By leveraging knowledge distillation from a teacher model and preference-guided optimization, Fast-ThinkAct efficiently transfers language and visual planning capabilities into embodied control policies. The approach maintains strong long-horizon planning, few-shot adaptation, and failure recovery abilities while substantially reducing computational overhead. Evaluated across multiple embodied manipulation and reasoning benchmarks, Fast-ThinkAct achieves up to an 89.3% reduction in inference latency without compromising task success rates or generalization performance, consistently outperforming prior methods.

Technology Category

• Planning, Routing, and Scheduling: Planning with Language Models
• Humans and AI: Human-Aware Planning and Behavior Prediction
• Computer Vision: Language and Vision

Application Category

• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Vision-Language-Action (VLA) tasks require reasoning over complex visual scenes and executing adaptive actions in dynamic environments. While recent studies on reasoning VLAs show that explicit chain-of-thought (CoT) can improve generalization, they suffer from high inference latency due to lengthy reasoning traces. We propose Fast-ThinkAct, an efficient reasoning framework that achieves compact yet performant planning through verbalizable latent reasoning. Fast-ThinkAct learns to reason efficiently with latent CoTs by distilling from a teacher, driven by a preference-guided objective to align manipulation trajectories that transfers both linguistic and visual planning capabilities for embodied control. This enables reasoning-enhanced policy learning that effectively connects compact reasoning to action execution. Extensive experiments across diverse embodied manipulation and reasoning benchmarks demonstrate that Fast-ThinkAct achieves strong performance with up to 89.3\% reduced inference latency over state-of-the-art reasoning VLAs, while maintaining effective long-horizon planning, few-shot adaptation, and failure recovery.

Problem

• Vision-Language-Action
• reasoning latency
• chain-of-thought
• embodied reasoning
• inference efficiency

Innovation

• latent reasoning
• vision-language-action
• chain-of-thought distillation
• embodied control
• efficient planning