2 papers
cs.LG2025
LLM Probing with Contrastive Eigenproblems: Improving Understanding and Applicability of CCS
Stefan F. Schouten, Peter Bloem
Contrast-Consistent Search (CCS) is an unsupervised probing method able to test whether large language models represent binary features, such as sentence truth, in their internal a…
cs.CL2025
Truth-value judgment in language models: 'truth directions' are context sensitive
Stefan F. Schouten, Peter Bloem, Ilia Markov +1
Recent work has demonstrated that the latent spaces of large language models (LLMs) contain directions predictive of the truth of sentences. Multiple methods recover such direction…