Nous: An Attempt to Extract and Inject the Cognition Behind Prediction-Market Behavior

The paper presents "Nous," a framework designed to extract human cognitive diversity from prediction market behaviors and inject it into LLM agents via prompts. It identifies an eight-dimensional behavioral profile from Polymarket trading, demonstrating that while extraction partially succeeds—showing stability in several parameters and some correlation with future profits—prompt-level injection fails to enhance model performance, indicating limitations in addressing cognitive monoculture through mere prompt adjustments. This highlights the need for more robust methods like fine-tuning to effectively diversify LLM outputs in prediction markets.