Reproduce “Asking Less, Adapting Better: Feedback-Efficient Test-Time Adaptation for Vision-and-Language Navigation” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Zixu Wang, Huaquan Zhang, Shoukai Xu, Jinwu Hu, Zhiquan Wen, Mingkui Tan
FETA adapts vision-and-language navigation policies online in unfamiliar environments by selecting one action per episode for binary feedback and updating the policy locally. It filters trajectories by variation in normalized entropy, then selects the action with the most sensitive entropy gradient. With DUET on REVERIE validation-unseen, SPL reaches 50.55 versus ATENA’s 42.73, using simulator feedback rather than live human labels.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.