Dataset · openai
Long-context noisy fact-retrieval benchmark
openai/BrowseCompLongContext
A set of questions and URLs from OpenAI to test the ability of LLMs to find relevant information among noise in a long context.
What's inside
The dataset is a text-based question-answering benchmark adapted from the BrowseComp agent benchmark for a long-context scenario. It contains fewer than 1,000 examples: each question is accompanied by a list of URLs marked as either required for answering or distractor noise.
What you can do with it
- Evaluating long-context language model performance and robustness against informational noise
- Testing RAG systems and pipelines for extracting relevant data from a set of web documents
- Tasks
- Question answering
- Size
- up to 1K rows
- Downloads this month
- 6.3K
- Likes
- 57
Models trained on it
How to load it
from datasets import load_dataset
ds = load_dataset("openai/BrowseCompLongContext")