MIRAI: Evaluating LLM Agents for International Event Forecasting

We present MIRAI, a benchmark designed to systematically evaluate LLM agents as temporal forecasters to predict international events. Our benchmark features an agentic environment with tools for accessing an extensive database of historical, structured events and textual news articles. We refine the GDELT event database with careful cleaning and parsing to curate a series of relational prediction tasks with varying forecasting horizons, assessing LLM agents’ abilities from short-term to long-term forecasting. We further implement APIs to enable LLM agents to utilize different tools via a code-based interface. Notably, MIRAI features a dynamic data construction pipeline that supports periodically downloading recent news and events, and automatically generates the most recent test split. This allows us to evaluate any newly released model in a contamination-free manner as we can always construct a test split later than its knowledge cutoff date. MIRAI comprehensively evaluates the agents’ capabilities in three dimensions: 1) autonomously source and integrate critical information from large global databases; 2) write codes with both domain-specific APIs and libraries for tool-use; and 3) jointly reason over historical knowledge from diverse formats and timespan to accurately predict future events. Through comprehensive evaluation, we establish a reliable benchmark for assessing the capabilities of LLM agents in forecasting international events and contribute to the development of more accurate and trustworthy models for international relation analysis.