Evaluating Long-Form Forecasts by Their Effect on Downstream Predictions
Published in ICML 2026 AI Forecasting Workshop, 2026
We argue that the value of a long-form forecast lies in how it updates the world model of a downstream predictor. We measure how conditioning on a long-form forecast improves the prediction accuracy of a weaker model across a sample of real-world events, testing seven frontier models under this framework.
Recommended citation: Qin, J. et al. (2026). "Evaluating Long-Form Forecasts by Their Effect on Downstream Predictions." ICML 2026 AI Forecasting Workshop.
Download Paper