Skip to content

Dataset · SISAP-Challenges

SISAP 2025 Vector Search and Indexing Benchmark

SISAP-Challenges/SISAP2025

Ready-to-use embedding dataset for evaluating nearest neighbor search algorithms.

What's inside

Data for the SISAP 2025 indexing competition, containing between 100 million and 1 billion records. It includes precomputed vector representations (384-dimensional embeddings from the all-MiniLM-L6-v2 model for cosine similarity) based on CCNEWS (614,664 articles) and Google QA (GooAQ) corpora.

What you can do with it

  • Testing and comparing speed and accuracy of vector search algorithms (ANN indexes)
  • Benchmarking vector databases on real text embeddings
Size
100M–1B rows
Downloads this month
24.7K
Likes
0

How to load it

from datasets import load_dataset

ds = load_dataset("SISAP-Challenges/SISAP2025")