Skip to content

Dataset · allenai

Verifiable instruction following test benchmark

allenai/IFBench_test

A small text prompt dataset from the Allen Institute for AI for evaluating instruction-following accuracy.

What's inside

A test split of under 1,000 text examples prepared by Ai2 researchers for the paper on Generalizing Verifiable Instruction Following. The dataset is distributed under the ODC-BY-1.0 license and includes responses from third-party models.

What you can do with it

  • Evaluating LLM ability to strictly adhere to specified constraints and prompt formats
  • Comparing models against each other on a standardized benchmark from Ai2
Size
up to 1K rows
Downloads this month
29.2K
Likes
21

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("allenai/IFBench_test")