Reproduce “Context-aware Dynamics Model for Generalization in Model-Based 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
Kimin Lee, Younggyo Seo, Seunghyun Lee, Honglak Lee, Jinwoo Shin
CaDM conditions a dynamics model on a context vector inferred from recent transitions, training it with forward, backward, and multi-step prediction. In simulated control tasks with unseen but related dynamics, PE-TS plus CaDM raises HalfCheetah average return in the moderate test regime from 2,019.6 to 7,087.2 across five runs. This is one task and regime, not evidence of real-world reliability.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.