Xiaodan Zhu

The paper proposes using question-asking as an inference-time intervention to probe a language model's hidden state, finding that the self-diagnosis process provides a predictive signal for final correctness, though it notes a gap between detection and recovery.