Paper Overview
Field: Computer Vision Authors: Xinhao Liu, Yunsheng Tian, Yuheng Qiu, Pieter Abbeel, Xi Chen, Guanya Shi, Chen Feng Posted: 2026-10-16 arXiv: 2610.23841
Abstract
Pretrained VLMs bring rich world knowledge and semantic priors to object-goal navigation, and fine-tuning a VLM into a navigation policy has become a common recipe. Despite this progress, success rates remain low, and it is unclear which capability these fine-tuned policies lack.
Our diagnosis identifies goal grounding as the dominant failure. These policies explore well, yet often fail to notice that the goal is already in view or to estimate how far away it is. To address this failure, we ground the goal explicitly.
Explicit grounding substantially improves success and reaches state-of-the-art results on six benchmarks across three scene collections. We also correct annotation defects in these benchmarks, so that future methods are evaluated more faithfully.
Our analysis on the cleaned benchmarks shows that when using ground-truth grounding, policies nearly match the experts that generated their demonstrations, and the remaining headroom lies in perception rather than policy learning. Our work points to goal grounding as the missing piece in object-goal navigation and provides a simple yet effective solution for this task.
Key Takeaways
- Diagnosis: Fine-tuned VLM navigation policies struggle primarily with goal grounding — recognizing the goal in view and estimating its distance — not with exploration.
- Solution: Explicit goal grounding yields large success-rate improvements, achieving SOTA on six benchmarks across three scene collections.
- Benchmark cleanup: The authors fix annotation defects in existing benchmarks for more faithful evaluation.
- Insight: With ground-truth grounding, policies nearly match expert demonstrators, indicating future gains should come from perception rather than policy learning.