README.md

metadata

dataset_info:
  features:
    - name: IMAGE
      dtype: image
    - name: PROMPT
      dtype: string
    - name: ANSWER
      dtype: string
  splits:
    - name: train
      num_bytes: 515885
      num_examples: 38
  download_size: 492703
  dataset_size: 515885
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
license: apache-2.0
language:
  - en
tags:
  - prompts
  - vibe
pretty_name: Difficult prompts
size_categories:
  - 1K<n<10K

Difficult Prompts Dataset 🏋️‍♂️

This dataset contains a curated collection of challenging prompts, consisting of text-only, image-only, and image-text pairs. It is designed for comparing the performance of different AI models on complex and diverse inputs.

The prompts and images were originally extracted from responses to this tweet.

Feel free to explore and use this dataset for 👨‍💻 vibing and figuring out how models behave.