VLN-Cache: Enabling Token Caching for VLN Models with Visual/Semantic Dynamics Awareness

TL;DR

VLN-Cache introduces a dynamic-aware token caching method for vision-and-language navigation models, significantly reducing inference costs while maintaining navigation performance.

Contribution

It proposes view-aligned remapping and semantic relevance filtering to handle visual and semantic dynamics in token caching for VLN models.

Findings

Achieves up to 1.52x speedup in navigation tasks.

Maintains competitive success rates with reduced inference costs.

Effectively handles visual and semantic shifts during navigation.

Abstract

Vision-and-Language Navigation (VLN) increasingly relies on large vision-language models, but their inference cost conflicts with real-time deployment. Token caching is a promising training-free strategy that avoids redundant computation by reusing stable visual tokens across frames. However, existing methods assume a static camera and fixed semantic focus, assumptions that VLN fundamentally violates. We identify two failure modes: (1) visual dynamics, where viewpoint shift displaces token positions across frames, causing position-wise matching to pair misaligned content; (2) semantic dynamics, where token relevance shifts across task stages as navigation progresses, making cached states stale. We propose VLN-Cache, a visual-dynamic-aware and semantic-dynamic-aware caching framework that introduces view-aligned remapping to recover geometric correspondences and a task-relevance saliency filter to veto reuse at semantic transitions. A layer-adaptive entropy policy further balances the per-layer reuse budget. Experiments on the R2R-CE simulation benchmark show up to 1.52x speedup while maintaining competitive navigation success rates.