Reproduce “RAIL: Representation-Aligned Imitation Learning for Student-Compatible Teacher Policies” 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
Meraj Mammadov, Pedro Zuidberg Dos Martires, Johannes Andreas Stork
RAIL trains a teacher to act through representations built from information shared with a limited-view student. Separate encoders align paired observations, and one policy uses either representation while policy gradients cannot update the encoders. In TunnelVision, teacher and transferred student each reach 1.00 mean success rate and a 0.00 imitation gap; evidence covers four simulated environments, not a general guarantee.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.