
Language models can be fine-tuned to generate well-calibrated linguistic expressions of uncertainty that accurately reflect the likelihood of their predictions being correct.


coremsg

equipping-language-models-with-calibrated-linguistic-expressions-of-uncertainty


Equipping Language Models with Calibrated Linguistic Expressions of Uncertainty


title_rewrite


Uncertainty measures for language models should be evaluated based on their ability to accurately reflect the expected correctness of generated outputs, without relying on ad hoc thresholding of correctness scores.


assessing-uncertainty-measures-for-language-models-a-rank-calibration-approach


Assessing Uncertainty Measures for Language Models: A Rank-Calibration Approach
