Reproduce “Double Check Your State Before Trusting It: Confidence-Aware Bidirectional Offline Model-Based Imagination” 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
Jiafei Lyu, Xiu Li, Zongqing Lu
CABI augments fixed offline datasets for model-free reinforcement learning by generating transitions with forward and backward dynamics models, then retaining the 20% with smallest cross-model state deviations in each mini-batch. On Adroit, CABI+BCQ totals 607.0 across 12 task-dataset combinations, versus 529.2 without the filter; agreement is a consistency check, not proof that transitions are true or safe.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.