Xinxin Fan
Titles as the sources gave them. In 1 entry, with 0 checked quotations.
In the Base
Every entry they appear in.
Uncovering Why LLMs Fail: From Internal Mechanisms to Adversarial Reasoning Trajectories
According to the abstract, Ding, Fan, Yuan and co-authors present CauTrace, a white-box framework that looks at a model's internal states to explain why reasoning fails. They evaluate it on ordinary reasoning failures and on paired adversarial perturbations, in Qwen and LLaMA-family models.