💡 Large Feedback Dataset
The RLAIF-V-Dataset is a large multimodal recall dataset. The dataset is built using open source models to provide high quality feedback.
The RLAIF-V-Dataset is a novel method of using open source MLLM to provide high quality feedback from deconfined model responses. By training on this data, models can achieve a higher level of confidence than existing open source models.
Load a dataset using a Python script as follows:
from datasets import load_dataset
dataset = load_dataset("HaoyeZhang/RLAIF-V-Dataset")