Reproduce “Productive State-Space Exploration for Multi-turn RL in LLM Agents” 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
Mohammed Haroon Dupty, Yong Liang Goh, Song Guo, Wee Sun Lee
The authors introduce an episode-memory penalty that gives negative intrinsic reward when an agent’s new observation exactly repeats an earlier one, while novel observations receive no bonus and task reward remains. In a single-run ALFWorld ablation, success rose from 39.1% to 57.8% and duplicate observations fell from 59.3% to 37.0%, evidence limited to this setup.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.