Reproduce “Unsupervised Skill Discovery with Bottleneck Option 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

Jaekyeom Kim, Seohong Park, Gunhee Kim

IBOL discovers reusable skills for simulated robot control without task rewards or demonstrations. It first trains a goal-conditioned linearizer, then learns a bottleneck representation of trajectories and a skill policy to imitate goals. It leads three shared-linearizer baselines on three information measures across four environments and reports better results on four controller tasks; vision-based settings remain untested.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.