DART-VLN: Test-Time Memory Decay and Anti-Loop Regularization for Discrete Vision-Language Navigation
2026-07-01 • Robotics
RoboticsArtificial Intelligence
AI summaryⓘ
The authors study how navigation agents that use memory can get stuck or confused when trying to find their way using partial information. They propose DART-VLN, a method that improves navigation during testing without needing any retraining or changing the agent's brain. Their approach reduces old, less useful memory and discourages the agent from moving back and forth unnecessarily. Tests show this makes the agents navigate more efficiently and reliably on standard benchmarks. This shows that small changes during testing can help memory-based navigation agents perform better.
vision-language navigationpartial observabilitymemory readouttest-time controlmemory decayaction selectionbacktrackingfrozen backboneR2R datasetREVERIE dataset
Authors
Shaoheng Zhang, Zhichen Li, Jie Mei
Abstract
Memory-based discrete vision-language navigation (VLN) agents must act under partial observability, yet even strong frozen backbones remain vulnerable at test time. Two common failure modes are stale historical evidence at memory readout and inefficient local backtracking during action selection. We present DART-VLN, a training-free test-time control framework for discrete VLN. DART-VLN combines Test-Time Memory Decay, a read-side memory reweighting rule that suppresses stale and redundant evidence without rewriting stored content, with Anti-Loop Regularization, a lightweight next-hop penalty that discourages immediate reversals during action selection. The framework introduces no new learnable parameters and leaves the learned backbone unchanged. Experiments on R2R and REVERIE show a consistent pattern: decay-only provides stable read-side gains, while decay+anti-loop achieves the best overall quality-efficiency trade-off, yielding shorter trajectories, lower runtime, and improved navigation performance in key settings. Behavioral analysis further confirms that anti-loop regularization reduces local backtracking and improves path efficiency under frozen backbones. Overall, the results show that modest test-time control can make memory-based discrete VLN more reliable and efficient without retraining.