Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Huynh, Jessica, Gomez, Alfredo, Deviyani, Athiya, Shelby, Renee, Bigham, Jeffrey P., Diaz, Fernando
Format: Preprint
Veröffentlicht: 2026
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2605.06283
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866909021966958592
author Huynh, Jessica
Gomez, Alfredo
Deviyani, Athiya
Shelby, Renee
Bigham, Jeffrey P.
Diaz, Fernando
author_facet Huynh, Jessica
Gomez, Alfredo
Deviyani, Athiya
Shelby, Renee
Bigham, Jeffrey P.
Diaz, Fernando
contents Autoraters, also referred to as LLM-as-judges, are increasingly used for evaluation and automated content moderation. However, there is limited statistical analysis of how modifications in a rubric presented to both humans and autoraters affect their score agreement. Rubrics that ask for an overall or \emph{holistic} judgment - for example, rating the ``quality'' of an essay - may be inconsistently interpreted due to the complexity or subjectivity of the criteria. Conversely, rubrics can ask for \emph{analytic} judgments, which decompose assessment criteria - for example, ``quality'' into ``fluency'' and ``organization''. While these rubrics can be edited to improve the individual accuracy of both human and automated scoring, this approach may result in disagreement between the two scores, or with the associated holistic judgment. Designing and deploying reliable autoraters requires understanding not just the relationship between human and autorater annotations but how that relationship changes as holistic or analytic judgments are elicited. The results indicate that rubric edits providing representative examples and additional context, and reducing positional bias in the rubric increased human-autorater agreement, while higher rubric complexity and conservative aggregation methods tended to decrease it. The findings from the automatic essay scoring and instruction-following evaluation domains suggest that practitioners should carefully analyze domain- and rubric-specific performance to move towards higher human-autorater agreement.
format Preprint
id arxiv_https___arxiv_org_abs_2605_06283
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Quantifying the Statistical Effect of Rubric Modifications on Human-Autorater Agreement
Huynh, Jessica
Gomez, Alfredo
Deviyani, Athiya
Shelby, Renee
Bigham, Jeffrey P.
Diaz, Fernando
Computation and Language
Autoraters, also referred to as LLM-as-judges, are increasingly used for evaluation and automated content moderation. However, there is limited statistical analysis of how modifications in a rubric presented to both humans and autoraters affect their score agreement. Rubrics that ask for an overall or \emph{holistic} judgment - for example, rating the ``quality'' of an essay - may be inconsistently interpreted due to the complexity or subjectivity of the criteria. Conversely, rubrics can ask for \emph{analytic} judgments, which decompose assessment criteria - for example, ``quality'' into ``fluency'' and ``organization''. While these rubrics can be edited to improve the individual accuracy of both human and automated scoring, this approach may result in disagreement between the two scores, or with the associated holistic judgment. Designing and deploying reliable autoraters requires understanding not just the relationship between human and autorater annotations but how that relationship changes as holistic or analytic judgments are elicited. The results indicate that rubric edits providing representative examples and additional context, and reducing positional bias in the rubric increased human-autorater agreement, while higher rubric complexity and conservative aggregation methods tended to decrease it. The findings from the automatic essay scoring and instruction-following evaluation domains suggest that practitioners should carefully analyze domain- and rubric-specific performance to move towards higher human-autorater agreement.
title Quantifying the Statistical Effect of Rubric Modifications on Human-Autorater Agreement
topic Computation and Language
url https://arxiv.org/abs/2605.06283