Reproduce “Inferring Transferable Rewards via Active Inverse Reinforcement Learning” 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
Victor Villin, Till Freihaut, Andreas Schlaginhaufen, Christos Dimitrakakis, Maryam Kamgarpour, Giorgia Ramponi
The authors present an active environment selection method for inverse reinforcement learning that chooses dynamics to identify reward components needed for transfer. It first cycles through available dynamics, then selects those with high estimated information gain. In the maze task, active-only reaches at 20 trajectories the suboptimality round-robin reaches at 100. This result is task-specific; the two-phase method catches up after about 60 trajectories.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.