Skip to content

Dataset · Fancy-MLLM

Dataset for developing multimodal reasoning

Fancy-MLLM/R1-Onevision

Set of complex multi-step reasoning tasks at the intersection of visual and textual understanding.

What's inside

Contains 100,000 to 1 million question-answering records (text modality specified in metadata, language unspecified). Covers reasoning tasks in natural scenes, science, and mathematics, connecting visual context with text analysis.

What you can do with it

  • SFT fine-tuning of multimodal models (MLLMs) for deep reasoning skills
  • Improving model accuracy in scientific and mathematical tasks with visual context
  • Evaluating model chain-of-thought reasoning capabilities on a specialized benchmark
Tasks
Question answering
Size
100K–1M rows
Downloads this month
3.2K
Likes
47

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("Fancy-MLLM/R1-Onevision")