Reproduce “AlphaRLM: Aligning Language Models for Literature-Grounded Symbolic Factor Discovery” 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

Jie Huang, Yuecheng Hong, Yao Chen, Hao Zhong, Linfeng Xiao, Xudong Liu, Qiang Ye, Hao Xia and 1 more

AlphaRLM builds stock-ranking signals by prompting a language model with papers and market data, then training its generation policy with rewards for executable formulas and separation in next-day returns among factor groups. In a backtest, its Qwen3-8B system achieves 6.94% annualized excess return on the S&P 500, versus 3.16% for the best non-AlphaRLM baseline; these are historical results, not live-market evidence.

Summary by alphaXiv.

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