Skip to content

Dataset · zai-org

LongBench v2 long-context benchmark

zai-org/LongBench-v2

An English test set to evaluate how LLMs handle complex long-context tasks.

What's inside

The repository contains English text tasks for deep analysis and reasoning in long contexts starting from 8k tokens and above. The dataset includes multiple-choice tasks, text classification, as well as questions on texts and tables. The size of the set is less than 1,000 examples, created by authors from THUDM.

What you can do with it

  • Testing LLMs' ability to retain and analyze long contexts
  • Comparing models in QA tasks and logical reasoning over large documents and tables
Tasks
Multiple choice, Question answering, Text classification, Table question answering
Languages
English
Size
up to 1K rows
Downloads this month
61K
Likes
65

How to load it

from datasets import load_dataset

ds = load_dataset("zai-org/LongBench-v2")