Learning to Judge: LLMs Designing and Applying Evaluation Rubrics

2026-02-09

AI Summary

This work addresses the limitations of traditional human-defined evaluation criteria, which are static and misaligned with large language models’ (LLMs) intrinsic understanding of linguistic quality. The authors propose leveraging LLMs to autonomously generate interpretable, task-specific evaluation dimensions and present the first systematic validation of LLMs’ capacity to construct such criteria. Using the GER-Eval framework with both closed-source models (e.g., GPT-4o) and open-source models (e.g., Llama), experiments demonstrate that LLMs can reliably produce semantically coherent evaluation standards, excelling particularly in non-factual tasks. Closed-source models significantly outperform open-source counterparts in cross-model generalization and alignment with human judgments, revealing high internal consistency within individual LLMs but notable fragmentation across different models.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Interpretability, Analysis, and Evaluation of NLP Models
• Search and Optimization: Evaluation and Analysis

Application Category

• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work

Abstract

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.

Problem

• evaluation rubrics
• large language models
• natural language generation
• human-model alignment
• scoring reliability

Innovation

• LLM evaluation
• evaluation rubrics
• GER-Eval
• scoring reliability
• cross-model generalization