Calibrating Verbal Uncertainty as a Linear Feature to Reduce Hallucinations

LLMs often adopt an assertive language style also when making false claims. Such "overconfident hallucinations" mislead users and erode trust. Achieving the ability to express in language the actual degree of uncertainty around a claim is therefore of great importance. We find that "verbal 1 uncertainty" is governed by a single linear feature in the representation space of LLMs, and show that this has only moderate correlation with the actual "semantic uncertainty" of the model. We apply this insight and show that (1) the mismatch between semantic and verbal uncertainty is a better predictor of hallucinations than semantic uncertainty alone and (2) we can intervene on verbal uncertainty at inference time and reduce confident hallucinations on short-form answers, achieving an average relative reduction of 30%. 2 * Equal Contribution † Work done during Internship at Meta FAIR 1 In this paper, we employ the term 'verbal' to mean 'pertaining to words rather than meaning or substance,' as opposed to 'spoken rather than written' (refer to Merriam-Webster's definitions: https://www.merriam-webster. com/dictionary/verbal ). Readers may substitute 'verbal uncertainty' with 'expressed uncertainty' throughout the text if they find it preferable.