Skip to content

Dataset · Lruca

Large video instruction dataset for multimodal tasks

Lruca/VidUEU100k

100k labeled video examples for SFT of multimodal models in video understanding and editing.

What's inside

This is a large-scale instruction-tuning dataset for multimodal models, containing videos and synthesized text tasks. The corpus was automatically collected using a specialized AI agent, VidUEU-Agent, which converts raw videos into metadata and generates instructions from them. The data accompanies the research paper of the same name (arXiv:2611.04418).

What you can do with it

  • Instruction-tuning (SFT) of multimodal video models (Video-LLM)
  • Training models for text-prompted video editing
  • Testing systems on general video understanding tasks
Downloads this month
406
Likes
5

How to load it

from datasets import load_dataset

ds = load_dataset("Lruca/VidUEU100k")