Reproduce “More: Mode-Optimized Reasoning with Self-Adaptive Effort for Self-Evolving 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
Dianxing Shi, Dingjie Fu, Tianyu Lei, Wangjie Qiu
MORE trains a self-evolving reasoner to choose among direct answers, short or long chain-of-thought, and code, while adapting effort to verifier-informed task difficulty. Unlike a fixed reasoning format or token budget, it learns both response form and effort. Across reported evaluations, MORE improves on corresponding AZR aggregate measures from 3B to 14B; this does not mean it wins every benchmark.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.