Kristjan Greenewald

Revisiting Group Relative Policy Optimization: Insights into On-Policy and Off-Policy Training

Stop Guessing When to Stop Testing: Efficient Model Evaluation with Just Enough Data

Differentially Private Wasserstein Barycenters

An Efficient and Modular Framework for Targeted Harm Mitigation in LLMs

Know What You Don't Know: Uncertainty Calibration of Process Reward Models

Compress then Serve: Serving Thousands of LoRA Adapters with Little Overhead

Activated LoRA: Fine-tuned LLMs for Intrinsics

Partially Observed Trajectory Inference using Optimal Transport and a Dynamics Prior

Compress then Serve: Serving Thousands of LoRA Adapters with Little Overhead