Paper

Improving Confidence Estimation in Language Models with Linear Representations of Ambiguity · ByteTrail