Saved in:
Bibliographic Details
Main Authors: Lemesle, Quentin, Jourdan, Léane, Munson, Daisy, Alain, Pierre, Chevelu, Jonathan, Delhay, Arnaud, Lolive, Damien
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2602.15778
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910271138693120
author Lemesle, Quentin
Jourdan, Léane
Munson, Daisy
Alain, Pierre
Chevelu, Jonathan
Delhay, Arnaud
Lolive, Damien
author_facet Lemesle, Quentin
Jourdan, Léane
Munson, Daisy
Alain, Pierre
Chevelu, Jonathan
Delhay, Arnaud
Lolive, Damien
contents Evaluating the quality of automatically generated text often relies on LLM-as-a-judge (LLM-judge) methods. While effective, these approaches are computationally expensive and require post-processing. To address these limitations, we build upon ParaPLUIE, a perplexity-based LLM-judge metric that estimates confidence over ``Yes/No'' answers without generating text. We introduce *-PLUIE, task specific prompting variants of ParaPLUIE and evaluate their alignment with human judgement. Our experiments show that personalised *-PLUIE achieves stronger correlations with human ratings while maintaining low computational cost.
format Preprint
id arxiv_https___arxiv_org_abs_2602_15778
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle *-PLUIE: Personalisable metric with Llm Used for Improved Evaluation
Lemesle, Quentin
Jourdan, Léane
Munson, Daisy
Alain, Pierre
Chevelu, Jonathan
Delhay, Arnaud
Lolive, Damien
Computation and Language
Evaluating the quality of automatically generated text often relies on LLM-as-a-judge (LLM-judge) methods. While effective, these approaches are computationally expensive and require post-processing. To address these limitations, we build upon ParaPLUIE, a perplexity-based LLM-judge metric that estimates confidence over ``Yes/No'' answers without generating text. We introduce *-PLUIE, task specific prompting variants of ParaPLUIE and evaluate their alignment with human judgement. Our experiments show that personalised *-PLUIE achieves stronger correlations with human ratings while maintaining low computational cost.
title *-PLUIE: Personalisable metric with Llm Used for Improved Evaluation
topic Computation and Language
url https://arxiv.org/abs/2602.15778