Skip to content

Dataset · Tevatron

Deep-Research benchmark for deep search systems

Tevatron/browsecomp-plus

Test set of complex search queries and a fixed corpus of 100 thousand web documents for fair agent comparison.

What's inside

The dataset contains fewer than 1000 complex logical reasoning queries from OpenAI's BrowseComp benchmark. Instead of searching the live internet, evaluation is conducted on a fixed corpus of approximately 100 thousand selected web documents, including human-verified evidence. This isolates the quality of the retriever and the LLM agent.

What you can do with it

  • Comparing and evaluating Deep-Research agents under reproducible conditions
  • Testing retrievers on complex multi-step queries
  • Debugging search and response synthesis pipelines over a large document base
Tasks
Question answering
Size
up to 1K rows
Downloads this month
33.7K
Likes
41

How to load it

from datasets import load_dataset

ds = load_dataset("Tevatron/browsecomp-plus")