PhD student, Massachusetts Institute of Technology
1 paper at NeurIPS 2025
We introduce a Particle Filtering approach to Inference Scaling that is robust against inherently imperfect reward models and performs significantly better than previous methods.
