Geometric Deviation as an Unsupervised Pre-Generation Reliability Signal: Probing LLM Representations for Answerability
Paper • 2605.03196 • Published
Yucheng Du
yucheng-du
·
AI & ML interests
LLM reliability; mechanistic interpretability; AI auditing; computational linguistics
Recent Activity
authored a paper about 4 hours ago
Geometric Deviation as an Unsupervised Pre-Generation Reliability Signal: Probing LLM Representations for Answerability authored a paper about 4 hours ago
Recognition-Refusal Misalignment in LLMs: Why Models Answer Structurally Unanswerable QuestionsOrganizations
authored 2
papers about 4 hours ago