AI-secure/adversarial-glue

[NeurIPS 2021] "Adversarial GLUE: A Multi-Task Benchmark for Robustness Evaluation of Language Models" by Boxin Wang*, Chejian Xu*, Shuohang Wang, Zhe Gan, Yu Cheng, Jianfeng Gao, Ahmed Hassan Awadallah, Bo Li.

24
/ 100
Experimental

This project helps machine learning researchers and natural language processing engineers rigorously test their language models against adversarial attacks. It provides a specialized dataset of text inputs designed to challenge a model's robustness. Researchers can input their language model's predictions on this dataset and receive official scores on both development and hidden test sets, allowing them to benchmark their model's resilience.

No commits in the last 6 months.

Use this if you are developing or evaluating language models and need to assess how robustly they perform when faced with subtly altered or challenging text inputs.

Not ideal if you are looking for a tool to train a language model from scratch or to apply existing models to standard text classification tasks.

natural-language-processing machine-learning-evaluation model-robustness text-classification AI-safety
No License Stale 6m No Package No Dependents
Maintenance 0 / 25
Adoption 5 / 25
Maturity 8 / 25
Community 11 / 25

How are scores calculated?

Stars

13

Forks

2

Language

Python

License

Last pushed

Apr 03, 2023

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/ml-frameworks/AI-secure/adversarial-glue"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.