lm_polygraph.estimators.boostedprob_score module

class lm_polygraph.estimators.boostedprob_score.BoostedProbSequence[source]

Bases: Estimator

Estimates the sequence-level uncertainty of a language model by taking the average of the token-level scores obtained with BoostedProb (https://aclanthology.org/2025.emnlp-main.166.pdf)