FanToMAS

We designed this game to incorporate human evaluation into the Identity Identification task. The task involves 20 characters with varying personality types, both fictional and non-fictional, drawn from diverse categories such as movies, politics, sports, and more. These characters were role-played by large language models (LLMs) to discuss a range of topics. While LLMs were used to evaluate the role-played responses, this game leverages human participants to further assess the quality of these evaluations.