Hejie Cui

Training a small advisor to steer frozen frontier LLMs actually degrades if it learns from every post-hoc correction—selectively distilling only the advice that causally shifts execution yields up to a 6.4% gain and robust cross-family model transfer.

Simulating 8.3 billion diverse personas reveals nuanced user interactions that traditional evaluations miss, transforming how we assess AI systems.

Multi-turn RL agents can learn far more effectively by explicitly monitoring and controlling uncertainty at both the token and turn levels, leading to more stable training and higher performance.