Skip to content

Dataset · skylenage-ai

Multimodal math problems with images

skylenage-ai/DeepVision-103K

100k to 1M visually diverse English math problems for training VLM reasoning.

What's inside

DeepVision-103K combines images and texts in English (the card indicates a range of 100K-1M examples) to train multimodal mathematical reasoning. The dataset features a wide range of topics, visual diversity, and answer verification capabilities. According to the card, training on it increases the accuracy and reasoning efficiency of models like Qwen3-VL-8B.

What you can do with it

  • SFT fine-tuning of multimodal models (VLM) for solving math and logic problems using images
  • Reducing reasoning token length while maintaining answer accuracy
Tasks
Image text → text
Languages
English
Size
100K–1M rows
Downloads this month
947
Likes
37

How to load it

from datasets import load_dataset

ds = load_dataset("skylenage-ai/DeepVision-103K")