Reproduce “Train An Imaginative Agent” 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

Jiaxin Ge, Junyi Zhang, Tianning Chai, Hanjun Yoo, Chenyang Wang, Xiuyu Li, Weijia Shi, XuDong Wang and 4 more

Agima uses one multimodal model to propose actions, predict resulting images, and feed predictions back as observations. On an 8×8 Pacman test with four food items, it succeeds in 54% of episodes versus 22% for linear rollout, at K = infinity, with no intermediate real observations. Evidence is limited to four benchmark environments.

Summary by alphaXiv.

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