Dataset · facebook
Complex adversarial natural language benchmark
facebook/anli
An English dataset for natural language inference (NLI) built with human and model participation.
What's inside
A large-scale English dataset (100K to 1M examples) for the natural language inference (NLI) task. Collected via an iterative adversarial method involving humans and neural networks across three rounds, each featuring train, dev, and test splits. Designed by the authors to be significantly harder than the classic SNLI and MNLI benchmarks.
What you can do with it
- Evaluating language model robustness against complex logical traps
- Fine-tuning classifiers for fact-checking and text contradiction tasks
- Comparing LLMs on a difficult NLI benchmark from the Facebook team
- Tasks
- Text classification
- Languages
- English
- Size
- 100K–1M rows
- Downloads this month
- 31.1K
- Likes
- 59
Models trained on it
- MoritzLaurer/mDeBERTa-v3-base-xnli-multilingual-nli-2mil7Zero shot classification 714.4K
- MoritzLaurer/DeBERTa-v3-base-mnli-fever-anliZero shot classification 197.3K
- MoritzLaurer/DeBERTa-v3-large-mnli-fever-anli-ling-wanliZero shot classification 96.3K
- tasksource/ModernBERT-base-nliZero shot classification 43.8K
- MoritzLaurer/DeBERTa-v3-xsmall-mnli-fever-anli-ling-binaryZero shot classification 9.4K
- tasksource/deberta-small-long-nliZero shot classification 9.3K
- tasksource/ModernBERT-large-nliZero shot classification 8.5K
- vicgalle/xlm-roberta-large-xnli-anliZero shot classification 1.9K
How to load it
from datasets import load_dataset
ds = load_dataset("facebook/anli")