Tevatron/browsecomp-plus

Name: Tevatron/browsecomp-plus
Creator: Tevatron
License: mit
Keywords: huggingface, task_categories:question-answering, license:mit, size_categories:n<1K, format:parquet, modality:text, library:datasets, library:dask, library:mlcroissant, question-answering

Question Answering · Tevatron· 58.2K

mit 2.6 GB

BrowseComp-Plus BrowseComp-Plus is a new benchmark for Deep-Research system, isolating the effect of the retriever and the LLM agent to enable fair, transparent comparisons of Deep-Research agents. The benchmark sources challenging, reasoning-intensive queries from OpenAI's BrowseComp. However, instead of searching the live web, BrowseComp-Plus evaluates against a fixed, curated corpus of ~100K web documents from the web. The corpus includes both human-verified evidence documents… See the full description on the dataset page: https://huggingface.co/datasets/Tevatron/browsecomp-plus.

Open in MLForge Sign up free Desktop app

# download instantly
mlforge datasets pull Tevatron/browsecomp-plus

Dataset details

Task

Question Answering

License

mit

Size

2.6 GB

Rows / images

830

Classes

Creator

Tevatron

Downloads

58.2K

Source

huggingface_datasets

Updated

2025-12-20

About Tevatron/browsecomp-plus

BrowseComp-Plus is a new benchmark for Deep-Research system, isolating the effect of the retriever and the LLM agent to enable fair, transparent comparisons of Deep-Research agents. The benchmark sources challenging, reasoning-intensive queries from OpenAI's BrowseComp. However, instead of searching the live web, BrowseComp-Plus evaluates against a fixed, curated corpus of ~100K web documents from the web. The corpus includes both human-verified evidence documents sufficient to answer the queries, and mined hard negatives to keep the task challenging.