How Does Beam Search improve Span-Level Confidence Estimation in Generative Sequence Labeling?

TL;DR

This work tackles the problem of reliable per-span confidence estimation in generative sequence labeling. It introduces beam-search–based confidence estimators—Aggregated Span Probability (AggSpan), Aggregated Sequence Probability (AggSeq), and the adaptive AdaAggSeq—to leverage top-k outputs and improve calibration over the conventional top-1 probability baseline. Across six diverse datasets, AggSpan and AggSeq reduce calibration error (ECE) compared to Span, with AdaAggSeq offering robust gains when beam size is varied. The methods remain applicable to black-box models that provide sequence-level probabilities and have practical implications for downstream systems that require reliable confidence signals. Overall, the approach advances practical confidence estimation for generation-based labeling and suggests directions for integrating calibrated scores into real-world IE/IR pipelines.

Abstract

Sequence labeling is a core task in text understanding for IE/IR systems. Text generation models have increasingly become the go-to solution for such tasks (e.g., entity extraction and dialog slot filling). While most research has focused on the labeling accuracy, a key aspect -- of vital practical importance -- has slipped through the cracks: understanding model confidence. More specifically, we lack a principled understanding of how to reliably gauge the confidence of a model in its predictions for each labeled span. This paper aims to provide some empirical insights on estimating model confidence for generative sequence labeling. Most notably, we find that simply using the decoder's output probabilities \textbf{is not} the best in realizing well-calibrated confidence estimates. As verified over six public datasets of different tasks, we show that our proposed approach -- which leverages statistics from top-$k$ predictions by a beam search -- significantly reduces calibration errors of the predictions of a generative sequence labeling model.