Reproduce “Toward Planning Representations That Separate Task Structure from Visual Content” 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
Aarav Srivastava, Vishwas Sathish, Rajesh P. N. Rao
The authors evaluate visual robustness in fixed-goal ObjectNav across five simulated houses using controlled single-factor and cumulative changes to appearance while preserving navigation layouts and start poses. Mean success falls from 0.94 to 0.66 under surface-only changes, while other changes generally have smaller effects; this measures behavior, not internal representations, and is limited to these houses.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.