What Are We Measuring in NLG? A Meta-Analysis of Evaluation Trends 2020-2025

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Yang, Jing, Feldhus, Nils, Mohtaj, Salar, Hennig, Leonhard, Wang, Qianli, Metheniti, Eleni, Hakimov, Sherzod, Jakob, Charlott, Solopova, Veronika, Rieck, Konrad, Schlangen, David, Möller, Sebastian, Schmitt, Vera
Natura: Preprint
Pubblicazione: 2026
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866911723929206784
author Yang, Jing
Feldhus, Nils
Mohtaj, Salar
Hennig, Leonhard
Wang, Qianli
Metheniti, Eleni
Hakimov, Sherzod
Jakob, Charlott
Solopova, Veronika
Rieck, Konrad
Schlangen, David
Möller, Sebastian
Schmitt, Vera
author_facet Yang, Jing
Feldhus, Nils
Mohtaj, Salar
Hennig, Leonhard
Wang, Qianli
Metheniti, Eleni
Hakimov, Sherzod
Jakob, Charlott
Solopova, Veronika
Rieck, Konrad
Schlangen, David
Möller, Sebastian
Schmitt, Vera
contents As Natural Language Generation (NLG) dominates modern NLP, scalable evaluation remains a critical bottleneck. Consequently, LLM-as-a-judge (LaaJ) adoption has accelerated rapidly, appearing in more papers than human evaluation in 2025. This pivotal shift motivates a critical analysis of current evaluation practices. Overcoming the limits of rigid keyword filtering and manual review, we employ a multi-LLM information extraction pipeline to gather structured metadata from 14,171 papers across four major NLP conferences (2020-2025). Analyzing 3,334 filtered NLG papers, we identify three systemic challenges. (1) Metric inertia: despite the shift toward open-ended generation, legacy lexical metrics (BLEU, ROUGE) persist as primary indicators, typically used alongside rather than replaced by semantic alternatives. (2) Metric-criteria mapping problem: our paper-level co-occurrence data reveals that general-purpose automatic metrics are applied as broad proxies for quality, without specifying which dimension of text generation they are intended to evaluate. (3) Validation gap: LaaJ has grown rapidly without commensurate human validation (fewer than 8% of papers). Crucially, while LaaJ correlates with aggregate quality, alignment collapses on fine-grained criteria like fluency. To address these gaps, we distill our findings into a minimal Evaluation Checklist to guide metric selection, construct validity, and LaaJ deployment.
format Preprint
id arxiv_https___arxiv_org_abs_2601_07648
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle What Are We Measuring in NLG? A Meta-Analysis of Evaluation Trends 2020-2025
Yang, Jing
Feldhus, Nils
Mohtaj, Salar
Hennig, Leonhard
Wang, Qianli
Metheniti, Eleni
Hakimov, Sherzod
Jakob, Charlott
Solopova, Veronika
Rieck, Konrad
Schlangen, David
Möller, Sebastian
Schmitt, Vera
Computation and Language
As Natural Language Generation (NLG) dominates modern NLP, scalable evaluation remains a critical bottleneck. Consequently, LLM-as-a-judge (LaaJ) adoption has accelerated rapidly, appearing in more papers than human evaluation in 2025. This pivotal shift motivates a critical analysis of current evaluation practices. Overcoming the limits of rigid keyword filtering and manual review, we employ a multi-LLM information extraction pipeline to gather structured metadata from 14,171 papers across four major NLP conferences (2020-2025). Analyzing 3,334 filtered NLG papers, we identify three systemic challenges. (1) Metric inertia: despite the shift toward open-ended generation, legacy lexical metrics (BLEU, ROUGE) persist as primary indicators, typically used alongside rather than replaced by semantic alternatives. (2) Metric-criteria mapping problem: our paper-level co-occurrence data reveals that general-purpose automatic metrics are applied as broad proxies for quality, without specifying which dimension of text generation they are intended to evaluate. (3) Validation gap: LaaJ has grown rapidly without commensurate human validation (fewer than 8% of papers). Crucially, while LaaJ correlates with aggregate quality, alignment collapses on fine-grained criteria like fluency. To address these gaps, we distill our findings into a minimal Evaluation Checklist to guide metric selection, construct validity, and LaaJ deployment.
title What Are We Measuring in NLG? A Meta-Analysis of Evaluation Trends 2020-2025
topic Computation and Language
url https://arxiv.org/abs/2601.07648