Reproduce “Speaking the Navigator's Language: Trajectory-Grounded Instruction Translation for Frozen Aerial VLN Agents” 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
Xi Chen, Zhe Liu, Xiaogang Xu, Jiafei Xu, Chunyi Zhou, Yuan Su, Rui Zeng, Tianyu Du and 3 more
TGIT learns an instruction translator for a frozen aerial navigation agent, using outcomes to select executable commands. The method trains on hard cases where the agent succeeds with an original instruction but fails with a shorter, human-informed rewrite. On 203 OpenFly episodes, it raises success rate from 15.27% to 37.93%, but path efficiency remains below the original-instruction result.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.