Multicalibration for Confidence Scoring in LLMs

This paper proposes the use of "multicalibration" to yield interpretable and reliable confidence scores for outputs generated by large language models (LLMs). Multicalibration asks for calibration not just marginally, but simultaneously across various intersecting groupings of the data. We...

Ausführliche Beschreibung

Bibliographische Detailangaben

Veröffentlicht in:	arXiv.org 2024-04
Hauptverfasser:	Detommaso, Gianluca, Bertran, Martin, Fogliato, Riccardo, Roth, Aaron
Format:	Artikel
Sprache:	eng
Schlagworte:	Algorithms Annotations Calibration Clustering Large language models Questions
Online-Zugang:	Volltext
Tags:	Tag hinzufügen Keine Tags, Fügen Sie den ersten Tag hinzu!