Reproduce “Latent Imagination for Parameterized Action Spaces” 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
Atahan Cilan, İhsan Mert Muhacıroğlu, Ümit Can Bekar
HyLI trains a hybrid-action policy in imagined trajectories from a learned world model. Counterfactual mode evaluation supplies candidate modes with value-based learning signals, while success-window replay preserves rare successful experience. At deployment, the encoder, recurrent state update, and policy choose actions without model search. Across seven simulated tasks, HyLI matches or exceeds published model-free baselines and decides 67× faster than DLPA on geometric mean.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.