Dragonjinny/

FiFA-pickapic-v2

These subsets were produced by filtering the original Pick-a-Pic v2 dataset using FiFA, a data filtering algorithm proposed in the paper Automated Filtering of Human Feedback Data for Aligning Text-to-Image Diffusion Models.

The filtering process is based on three key metrics:

Preference Margin: Estimated using PickScore Text Quality: Estimated through LLM scoring Text Diversity: Estimated using K-NN distance… See the full description on the dataset page: https://huggingface.co/datasets/Dragonjinny/FiFA-pickapic-v2.