From Rubrics to Reliable Scores: Evidence-Grounded Text Evaluation with LLM Judges
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2026
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866914609247551488 |
|---|---|
| author | Hong, Yihan Yao, Huaiyuan Shen, Bolin Xu, Wanpeng Wei, Hua Dong, Yushun |
| author_facet | Hong, Yihan Yao, Huaiyuan Shen, Bolin Xu, Wanpeng Wei, Hua Dong, Yushun |
| contents | Rubric-based text evaluation increasingly uses large language models (LLMs) as scalable judges, but aligning frozen black-box models with human scoring standards remains challenging. We formulate this challenge as a criteria-transfer problem: the goal is not merely to prompt an LLM to assign a score, but to transfer human rubric intent into a stable, auditable, and human-aligned scoring protocol. We identify three recurring failure modes in LLM-based rubric scoring: rubric execution drift, unverifiable score attribution, and human-scale misalignment. To address these failure modes, we introduce Rulers, a three-stage inference-time framework for reliable, evidence-grounded rubric-based text evaluation. Rulers first converts a human rubric into a locked task-level specification, then executes the specification with structured checklist decisions, typed evidence grounding, and extractive quote verification when applicable, and finally applies post-hoc calibration to align model-derived signals with human score boundaries. Across four rubric-governed benchmarks covering essay scoring, summarization assessment, EFL writing evaluation, and structured-input text generation, Rulers achieves stronger human-score agreement in most evaluated settings across multiple frozen backbone models. Further analyses show that Rulers better matches empirical human score distributions, improves stability under semantically equivalent rubric perturbations, and benefits from each of its three components. These results suggest that reliable LLM judging requires fixed criteria, traceable evidence, and calibrated score interpretation rather than prompt phrasing alone. Our code is available at https://anonymous.4open.science/r/Rulers_0525-3328. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2601_08654 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | From Rubrics to Reliable Scores: Evidence-Grounded Text Evaluation with LLM Judges Hong, Yihan Yao, Huaiyuan Shen, Bolin Xu, Wanpeng Wei, Hua Dong, Yushun Computation and Language Artificial Intelligence Machine Learning Rubric-based text evaluation increasingly uses large language models (LLMs) as scalable judges, but aligning frozen black-box models with human scoring standards remains challenging. We formulate this challenge as a criteria-transfer problem: the goal is not merely to prompt an LLM to assign a score, but to transfer human rubric intent into a stable, auditable, and human-aligned scoring protocol. We identify three recurring failure modes in LLM-based rubric scoring: rubric execution drift, unverifiable score attribution, and human-scale misalignment. To address these failure modes, we introduce Rulers, a three-stage inference-time framework for reliable, evidence-grounded rubric-based text evaluation. Rulers first converts a human rubric into a locked task-level specification, then executes the specification with structured checklist decisions, typed evidence grounding, and extractive quote verification when applicable, and finally applies post-hoc calibration to align model-derived signals with human score boundaries. Across four rubric-governed benchmarks covering essay scoring, summarization assessment, EFL writing evaluation, and structured-input text generation, Rulers achieves stronger human-score agreement in most evaluated settings across multiple frozen backbone models. Further analyses show that Rulers better matches empirical human score distributions, improves stability under semantically equivalent rubric perturbations, and benefits from each of its three components. These results suggest that reliable LLM judging requires fixed criteria, traceable evidence, and calibrated score interpretation rather than prompt phrasing alone. Our code is available at https://anonymous.4open.science/r/Rulers_0525-3328. |
| title | From Rubrics to Reliable Scores: Evidence-Grounded Text Evaluation with LLM Judges |
| topic | Computation and Language Artificial Intelligence Machine Learning |
| url | https://arxiv.org/abs/2601.08654 |