Reproduce “RLiD: Reinforcement Learning in the Decision Plane for Language Agents and LLMs” 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

Ziyi Chen, Yan Zhang, Jianhui Wei, Jianing Yin, Daoan Zhang

RLiD trains language agents to select executable actions from a finite menu, rather than generate text for parsing. It reads observations, optional memory, and labeled choices, selecting an action in one forward pass. On ALFWorld’s 134 unseen games, the full route achieved 99.3% success in 12.70 allocated H100 GPU-hours, versus 86.6% for GiGPO in 17.34; these budgets are not latency.

Summary by alphaXiv.

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