The GEM Benchmark: Natural Language Generation, its Evaluation and Metrics

2021/02/02 by Sebastian Gehrmann, Gehrmann, Sebastian, Tosin Adewumi +115 · 52 citations
Computer Science · Engineering · #Artificial intelligence #Benchmark (surveying) #Computer science #Data science #Engineering #Machine learning #Natural Language Processing Techniques #Natural language #Natural language generation #Programming language #Set (abstract data type) #Systems engineering #Task (project management) #Text Readability and Simplification #Topic Modeling #cs.AI #cs.CL #cs.LG

paper · pdf · doi:10.48550/arxiv.2102.01672

published in arXiv (Cornell University) (Cornell University)

openalex publication_date 2021/02/02 · arxiv created 2021/04/01 · arxiv updated 2021/04/02 · openalex created_date 2025/10/10 · openalex updated_date 2026/08/05

Abstract

We introduce GEM, a living benchmark for natural language Generation (NLG), its Evaluation, and Metrics. Measuring progress in NLG relies on a constantly evolving ecosystem of automated metrics, datasets, and human evaluation standards. Due to this moving target, new models often still evaluate on divergent anglo-centric corpora with well-established, but flawed, metrics. This disconnect makes it challenging to identify the limitations of current models and opportunities for progress. Addressing this limitation, GEM provides an environment in which models can easily be applied to a wide set of tasks and in which evaluation strategies can be tested. Regular updates to the benchmark will help NLG research become more multilingual and evolve the challenge alongside models. This paper serves as the description of the data for which we are organizing a shared task at our ACL 2021 Workshop and to which we invite the entire NLG community to participate.

Citations

• AmbigQA: Answering Ambiguous Open-domain Questions
• Evaluation of Text Generation: A Survey
• Multilingual Denoising Pre-training for Neural Machine Translation
• Datasheets for Datasets

Cited by

• Distribution-Specific Curvature Control with Finite-Sample Guarantees for Open-Weight Safety
• Eval Factsheets: A Structured Framework for Documenting AI Evaluations
• LoRA on the Go: Instance-level Dynamic LoRA Selection and Merging
• From the Rock Floor to the Cloud: A Systematic Survey of State-of-the-Art NLP in Battery Life Cycle
• Sunflower: A New Approach To Expanding Coverage of African Languages in Large Language Models
• Reference-Free Rating of LLM Responses via Latent Information
• Struct-Bench: A Benchmark for Differentially Private Structured Text Generation
• Survey of NLU Benchmarks Diagnosing Linguistic Phenomena: Why not Standardize Diagnostics Benchmarks?
• Compression, Transduction, and Creation: A Unified Framework for Evaluating Natural Language Generation
• cushLEPOR: customising hLEPOR metric using Optuna for higher agreement with human judgments or pre-trained language model LaBSE
• The Benchmark Lottery
• SynthTextEval: Synthetic Text Data Generation and Evaluation for High-Stakes Domains
• Benchmarking Multimodal AutoML for Tabular Data with Text Fields
• Optimal Control for Transformer Architectures: Enhancing Generalization, Robustness and Efficiency
• Automatic Construction of Evaluation Suites for Natural Language Generation Datasets
• MATCHA: Matching Text via Contrastive Semantic Alignment
• A Discussion on Building Practical NLP Leaderboards: The Case of Machine\n Translation
• Datasets: A Community Library for Natural Language Processing

Related

• GEMv2: Multilingual NLG Benchmarking in a Single Line of Code
• Data-driven Natural Language Generation: Paving the Road to Success
• Repairing the Cracked Foundation: A Survey of Obstacles in Evaluation Practices for Generated Text
• Automatic Construction of Evaluation Suites for Natural Language Generation Datasets
• LUNA: A Framework for Language Understanding and Naturalness Assessment