一作ACL 2026Poster· Natural Language Processing
How Context Shapes Truth: Geometric Transformations of Statement-level Truth Representations in LLM s
Shivam Adarsh · Maria Maistro · Christina Lioma
摘要
Large Language Models (LLMs) often encode whether a statement is true as a vector in their residual stream activations. These vectors, also known as truth vectors, have been studied in prior work, however how they change when context is introduced remains unexplored. We study this question by mea...