Reproduce “When News Gets Old: Portfolio Allocation with Reinforcement Learning and Large Language Models” 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
Bolin Song, Surui Xiang
The authors compare SEC disclosure signals with or without report timing information, and filtering signals more than eight weeks old. At a four-week delivery delay, a direct LLM policy with filtering averaged 18.43% annualized net return, equally averaged across three annual tests after costs. This exceeded news-using policies but not the market-data-only RL policy; results are retrospective.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.