CARREGANDO O RADAR…
What People Almost Did: Evaluating LLM Social Simulations Beyond Behavioral Fit | Radar arXiv · portela.dev