Aux-Think: Exploring Reasoning Strategies for Data-Efficient Vision-Language Navigation
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2025
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866908590451720192 |
|---|---|
| author | Wang, Shuo Wang, Yongcai Li, Wanting Cai, Xudong Wang, Yucheng Chen, Maiyue Wang, Kaihui Su, Zhizhong Li, Deying Fan, Zhaoxin |
| author_facet | Wang, Shuo Wang, Yongcai Li, Wanting Cai, Xudong Wang, Yucheng Chen, Maiyue Wang, Kaihui Su, Zhizhong Li, Deying Fan, Zhaoxin |
| contents | 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. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2505_11886 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Aux-Think: Exploring Reasoning Strategies for Data-Efficient Vision-Language Navigation Wang, Shuo Wang, Yongcai Li, Wanting Cai, Xudong Wang, Yucheng Chen, Maiyue Wang, Kaihui Su, Zhizhong Li, Deying Fan, Zhaoxin Robotics 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. |
| title | Aux-Think: Exploring Reasoning Strategies for Data-Efficient Vision-Language Navigation |
| topic | Robotics |
| url | https://arxiv.org/abs/2505.11886 |