Learning to Judge: LLMs Designing and Applying Evaluation Rubrics

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Siro, Clemencia, Aliannejadi, Pourya, Aliannejadi, Mohammad
Format: Preprint
Veröffentlicht: 2026
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866912890807648256
author Siro, Clemencia
Aliannejadi, Pourya
Aliannejadi, Mohammad
author_facet Siro, Clemencia
Aliannejadi, Pourya
Aliannejadi, Mohammad
contents Large language models (LLMs) are increasingly used as evaluators for natural language generation, applying human-defined rubrics to assess system outputs. However, human rubrics are often static and misaligned with how models internally represent language quality. We introduce GER-Eval (Generating Evaluation Rubrics for Evaluation) to investigate whether LLMs can design and apply their own evaluation rubrics. We evaluate the semantic coherence and scoring reliability of LLM-defined criteria and their alignment with human criteria. LLMs reliably generate interpretable and task-aware evaluation dimensions and apply them consistently within models, but their scoring reliability degrades in factual and knowledge-intensive settings. Closed-source models such as GPT-4o achieve higher agreement and cross-model generalization than open-weight models such as Llama. Our findings position evaluation as a learned linguistic capability of LLMs, consistent within models but fragmented across them, and call for new methods that jointly model human and LLM evaluative language to improve reliability and interpretability.
format Preprint
id arxiv_https___arxiv_org_abs_2602_08672
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Learning to Judge: LLMs Designing and Applying Evaluation Rubrics
Siro, Clemencia
Aliannejadi, Pourya
Aliannejadi, Mohammad
Computation and Language
Machine Learning
Large language models (LLMs) are increasingly used as evaluators for natural language generation, applying human-defined rubrics to assess system outputs. However, human rubrics are often static and misaligned with how models internally represent language quality. We introduce GER-Eval (Generating Evaluation Rubrics for Evaluation) to investigate whether LLMs can design and apply their own evaluation rubrics. We evaluate the semantic coherence and scoring reliability of LLM-defined criteria and their alignment with human criteria. LLMs reliably generate interpretable and task-aware evaluation dimensions and apply them consistently within models, but their scoring reliability degrades in factual and knowledge-intensive settings. Closed-source models such as GPT-4o achieve higher agreement and cross-model generalization than open-weight models such as Llama. Our findings position evaluation as a learned linguistic capability of LLMs, consistent within models but fragmented across them, and call for new methods that jointly model human and LLM evaluative language to improve reliability and interpretability.
title Learning to Judge: LLMs Designing and Applying Evaluation Rubrics
topic Computation and Language
Machine Learning
url https://arxiv.org/abs/2602.08672