Non-Exchangeable Conformal Language Generation with Nearest Neighbors
This work addresses the problem of reliable uncertainty estimation for text generation systems, offering a post-hoc method that can be applied to arbitrary models without extra training, though it is incremental in extending existing conformal prediction techniques.
The paper tackles the challenge of quantifying uncertainty in text generation by extending conformal prediction to non-exchangeable settings, using nearest neighbors to provide token-level prediction sets with statistical guarantees. Experiments in machine translation and language modeling show improved generation quality and tighter prediction sets with good coverage.
Quantifying uncertainty in automatically generated text is important for letting humans check potential hallucinations and making systems more reliable. Conformal prediction is an attractive framework to provide predictions imbued with statistical guarantees, however, its application to text generation is challenging since any i.i.d. assumptions are not realistic. In this paper, we bridge this gap by leveraging recent results on non-exchangeable conformal prediction, which still ensures bounds on coverage. The result, non-exchangeable conformal nucleus sampling, is a novel extension of the conformal prediction framework to generation based on nearest neighbors. Our method can be used post-hoc for an arbitrary model without extra training and supplies token-level, calibrated prediction sets equipped with statistical guarantees. Experiments in machine translation and language modeling show encouraging results in generation quality. By also producing tighter prediction sets with good coverage, we thus give a more theoretically principled way to perform sampling with conformal guarantees.