Dataset · facebook
LAMA benchmark for testing language model knowledge
facebook/lama
English text dataset from Facebook Research for evaluating factual knowledge in pretrained LLMs.
What's inside
The dataset contains 10k to 100k English examples designed for probing and analyzing factual and commonsense knowledge embedded in language models. Prepared by Facebook Research, it is used for text retrieval and classification tasks.
What you can do with it
- Evaluating factual knowledge volume in pretrained LLMs without fine-tuning
- Comparing models in fact extraction and fill-in-the-blank tasks
- Tasks
- Text retrieval, Text classification
- Languages
- English
- Size
- 10K–100K rows
- Downloads this month
- 3.3K
- Likes
- 20
Models trained on it
How to load it
from datasets import load_dataset
ds = load_dataset("facebook/lama")