Skip to content

Benchmarks

Jump to bottom
Max Kuhmichel edited this page Sep 13, 2023 · 22 revisions

Natural Language and Text-Based

Basic Benchmarks

General Language Understanding Evaluation (GLUE)
Homepage Paper Github

General language capabilities and multipurpose benchmarks

Question answering

Stanford Question Answering Dataset (SQuAD)
Homepage
arxiv.org SQuAD 1 SQuAD 2

Natural Language Understanding

Natural Language Generation

GLGE: A New General Language Generation Evaluation Benchmark
Natural Language Generation, 24 Taks 3 Difficulties; contains MASS, BART, and Prophet-Net Baselines
arxiv.org 2021 Link
Microsoft Research, College of Computer Science Sichuan University, Dayiheng Liu, et al.
Public Repository and guide https://microsoft.github.io/glge/

BERTSCORE: EVALUATING TEXT GENERATION WITH BERT
automatic evaluation metric for text generation
Cite LinkPaperGithub

Common reasoning

SWAG: A Large-Scale Adversarial Dataset for Grounded Commonsense Inference
Situations With Adversarial Generations
Homepage Paper

HellaSwag: Can a Machine Really Finish Your Sentence?
Like SWAG but harder
Homepage 2019 Paper

Clone this wiki locally