PhD student, Massachusetts Institute of Technology
1 paper at NeurIPS 2025
We provide theoretical analysis for forward and reverse KL-regularized RLHF under multiple reference models.