Yucheng Du

yucheng-du

1

·

https://yucheng-du.github.io/

AI & ML interests

LLM reliability, mechanistic interpretability, AI auditing, computational linguistics

Recent Activity

submitted a paper 1 day ago

Recognition-Refusal Misalignment in LLMs: Why Models Answer Structurally Unanswerable Questions

authored a paper 2 days ago

Geometric Deviation as an Unsupervised Pre-Generation Reliability Signal: Probing LLM Representations for Answerability

authored a paper 2 days ago

Recognition-Refusal Misalignment in LLMs: Why Models Answer Structurally Unanswerable Questions

View all activity

Organizations

Papers 2

arxiv:2608.29109

arxiv:2605.03196

models 0

None public yet

datasets 0

None public yet