Aux-Think: Exploring Reasoning Strategies for Data-Efficient Vision-Language Navigation

TL;DR

This paper investigates reasoning strategies in vision-language navigation, identifies inference-time reasoning issues, and proposes Aux-Think, a framework that improves data efficiency and performance by training models with structured reasoning supervision.

Contribution

It introduces Aux-Think, a novel training framework that internalizes reasoning patterns for VLN, and provides the first Chain-of-Thought dataset for this task.

Findings

Inference-time reasoning degrades navigation accuracy.

Aux-Think reduces training effort significantly.

Aux-Think achieves state-of-the-art performance with less data.

Abstract

Vision-Language Navigation (VLN) is a critical task for developing embodied agents that can follow natural language instructions to navigate in complex real-world environments. Recent advances in VLN by large pretrained models have significantly improved generalization and instruction grounding compared to traditional approaches. However, the role of reasoning strategies in navigation-an action-centric, long-horizon task-remains underexplored, despite Chain-of-Thought (CoT) reasoning's demonstrated success in static tasks like visual question answering. To address this gap, we conduct the first systematic evaluation of reasoning strategies for VLN, including No-Think (direct action prediction), Pre-Think (reason before action), and Post-Think (reason after action). Surprisingly, our findings reveal the Inference-time Reasoning Collapse issue, where inference-time reasoning degrades navigation accuracy, highlighting the challenges of integrating reasoning into VLN. Based on this insight, we propose Aux-Think, a framework that trains models to internalize structured reasoning patterns through CoT supervision, while inferring action directly without reasoning in online prediction. To support this framework, we release R2R-CoT-320k, the first Chain-of-Thought annotated dataset for VLN. Extensive experiments show that Aux-Think reduces training effort greatly and achieves the best performance under the same data scale.