A study investigating whether format restrictions like JSON or XML impact the performance of large language models (LLMs) in tasks like reasoning and domain knowledge comprehension.
A new model developed by researchers at MIT and the University of Washington predicts human goals or actions more accurately than previous models. The latent inference budget model identifies patterns in human or machine decision-making and uses this information to forecast behavior.
