ArXiv▲ 43

One MLLM, One Call: Efficient Zero-Shot Vision-and-Language Navigation via Spatial-Aware Waypoints

ArXiv|Sunday, September 6, 2026|7 days ago|Shiqi Pan

No AI summary available for this article.

Why It Matters

Vision-and-Language Navigation in Continuous Environments (VLN-CE) requires an embodied agent to navigate unseen environments by following natural language instructions.

Provenance

Discovered via ArXiv and published by ArXiv.

Open original ↗Source profile →

Key Claims

Current zero-shot VLN-CE methods either rely on pre-trained waypoint predictors or require multiple queries to large models per step.
To address prohibitive inference latency and computational overhead, we propose O2C-Nav, an efficient zero-shot navigation framework that calls only a single large model once per decision step.
Our approach introduces a training-free structured waypoint generator and a novel abstract representat...

Original description

Vision-and-Language Navigation in Continuous Environments (VLN-CE) requires an embodied agent to navigate unseen environments by following natural language instructions. Current zero-shot VLN-CE methods either rely on pre-trained waypoint predictors or require multiple queries to large models per step. To address prohibitive inference latency and computational overhead, we propose O2C-Nav, an efficient zero-shot navigation framework that calls only a single large model once per decision step. Our approach introduces a training-free structured waypoint generator and a novel abstract representat...

arxiv cs.CV cs.AI publisher:arxiv tier:research

Read on ArXiv ↗← Back to feed

Discovered via ArXiv

Research papers and preprints from arXiv.

Publisher: arxiv.org

ID: http://arxiv.org/abs/2609.06476v1 · Indexed 4 days ago