Dataset · Xkev
100k VQA examples with step-by-step reasoning
Xkev/LLaVA-CoT-100k
VQA data with structured reasoning annotations for training Vision-Language Models.
What's inside
A dataset from the LLaVA-CoT paper: visual question-answering examples collected from various sources with structured reasoning annotations. It is needed so that Vision-Language Models learn autonomous multi-step reasoning for complex visual questions. Language is English, size is from 10K to 100K examples.
What you can do with it
- SFT Vision-Language Model for step-by-step visual question answering
- LoRA for LLaVA-like model on structured reasoning
- Evaluation of multi-step visual reasoning of VLMs
- Tasks
- Visual question answering, Image text → text
- Languages
- English
- Size
- 10K–100K rows
- Downloads this month
- 3.5K
- Likes
- 109
Models trained on it
- Xkev/Llama-3.2V-11B-cotImage text → text 777
- Tuwhy/Llama-3.2V-11B-Sherlock-iter1Image text → text 72
- omkarthawakar/LlamaV-o1Question answering 51
- BarraHome/Mistroll-3.0-CoT-Llama-3.2-11B-Vision-InstructImage text → text 32
- Tuwhy/Llama-3.2V-11B-Sherlock-iter2Image text → text 32
- Tuwhy/Llama-3.2V-11B-Sherlock-OfflineImage text → text 28
- Tuwhy/Llama-3.2V-11B-Sherlock-SFTImage text → text 24
- huyquoctrinh/FireboltVL 14
How to load it
from datasets import load_dataset
ds = load_dataset("Xkev/LLaVA-CoT-100k")