Research
Formal papers, empirical results, datasets, and technical methodology. Our work bridges academic rigor with practical market infrastructure.
LLM as a Risk Manager: LLM Semantic Filtering for Lead-Lag Trading in Prediction Markets
Granger causality discovers candidate leader-follower pairs across prediction markets; an LLM then filters for relationships with plausible economic transmission mechanisms. The hybrid strategy improved win rates from 51.4% to 54.5% and cut average losses by 46.5%.
River summary, results, and downloadsForecasting Future Language: Context Design for Mention Markets
How should an LLM forecast whether a company will say a specific word on its earnings call? Market-Conditioned Prompting treats the market-implied probability as a prior and updates it with transcripts and news, beating both the model alone and plain-context prompting.
River summary, results, and downloads