|
|
--- |
|
|
language: |
|
|
- en |
|
|
- zh |
|
|
size_categories: |
|
|
- n<1K |
|
|
task_categories: |
|
|
- visual-question-answering |
|
|
tags: |
|
|
- multimodal |
|
|
- LLM |
|
|
- MLLM |
|
|
--- |
|
|
|
|
|
### Oasis: One Image is All You Need for Multimodal Instruction Data Synthesis |
|
|
This dataset contains **Oasis-500k** dataset. |
|
|
|
|
|
<a href="https://huggingface.co/papers/2503.08741">[Read the Paper]</a> | <a href="https://github.com/Letian2003/MM_INF">[Github Repo]</a> |
|
|
|
|
|
 |
|
|
|
|
|
All images come from Cambrian-10M. Instructions and responses are generated by MLLM. |
