1 paper
Benjamin Turtel, Paul Wilczewski, Kris Skotheim +2
This paper evaluates how reward function choice shapes the performance and behavior of LLM forecasters. We compare five proper scoring rules as training objectives for binary forec…