Hua Wei

LLMs have "pure incorrectness" features that correlate with wrong answers but don't actually *cause* them, suggesting that simply identifying error-correlated activations isn't enough for effective intervention.

LLM agents can learn to cooperate far more efficiently by borrowing credit assignment techniques from classic multi-agent RL.

LLMs learn faster and perform better in decision-making tasks when rewarded for being uncertain, not just for succeeding.

Forget weighting preferences alone – this new method uses conformal prediction to directly quantify and leverage the reliability of the *answers* themselves, leading to more robust and data-efficient LLM alignment.