Arena.ai, the company behind community LLM evaluations, announced it chose seven projects for the Spring 2026 cohort. Named recipients include Emmanuel Candès at Stanford for work on pairwise human preference data and Haifeng Xu at the University of Chicago for studies on predictive capabilities. The selection ties to Arena's Academic Partnerships Program, which invites proposals from U.S. tenure-track faculty on topics such as AI evaluation, measurement, human preference, safety, and alignment.