OneBench
Confident at the moment of action: belief miscalibration in LLM play under hidden information | OneBench: AI Insights