Dataset · mteb
SemEval STS 2013 semantic textual similarity benchmark
mteb/sts13-sts
English text pairs from web and news for evaluating embedding models in MTEB.
What's inside
The classic SemEval STS 2013 dataset adapted for the Massive Text Embedding Benchmark (MTEB). Contains 1k to 10k English text examples from news, web, and popular science sources for evaluating sentence similarity. Supports running evaluation directly via the mteb package.
What you can do with it
- Testing the quality of universal text embeddings on semantic similarity tasks
- Including in validation pipelines when training or quantizing encoder language models
- Tasks
- Sentence similarity
- Languages
- English
- Size
- 1K–10K rows
- Downloads this month
- 20.7K
- Likes
- 2
How to load it
from datasets import load_dataset
ds = load_dataset("mteb/sts13-sts")