VLA-Thinker: Boosting Vision-Language-Action Models through Thinking-with-Image Reasoning

TL;DR

VLA-Thinker introduces a dynamic reasoning framework for vision-language-action models, enabling active environment revisiting and improved long-horizon task performance through a two-stage training pipeline.

Contribution

It presents a novel thinking-with-image reasoning approach with a specialized training pipeline, enhancing embodied intelligence in VLA models.

Findings

Achieves 97.5% success on LIBERO benchmark.

Significantly improves manipulation performance.

Demonstrates strong gains on long-horizon robotic tasks.

Abstract

Vision-Language-Action (VLA) models have shown promising capabilities for embodied intelligence, but most existing approaches rely on text-based chain-of-thought reasoning where visual inputs are treated as static context. This limits the ability of the model to actively revisit the environment and resolve ambiguities during long-horizon tasks. We propose VLA-Thinker, a thinking-with-image reasoning framework that models perception as a dynamically invocable reasoning action. To train such a system, we introduce a two-stage training pipeline consisting of (1) an SFT cold-start phase with curated visual Chain-of-Thought data to activate structured reasoning and tool-use behaviors, and (2) GRPO-based reinforcement learning to align complete reasoning-action trajectories with task-level success. Extensive experiments on LIBERO and RoboTwin 2.0 benchmarks demonstrate that VLA-Thinker significantly improves manipulation performance, achieving 97.5% success rate on LIBERO and strong gains across long-horizon robotic tasks. Project and Codes: https://cywang735.github.io/VLA-Thinker/ .