Starrrrrry/

RLAIF-V-Bias-Dataset

🚀 Project Page: https://zhangzef.github.io/NaPO-Project-Page/ The RLAIF-V-Bias-Dataset is constructed based on the RLAIF-V-Dataset to mitigate the issue of modality bias in MLLMs using the LLaVA-v1.5-7b model.

RLAIF-V-Dataset provides high-quality feedback with a total number of 83,132 preference pairs, where the instructions are collected from a diverse range of datasets including MSCOCO, ShareGPT-4V, MovieNet, Google Landmark v2, VQA v2… See the full description on the dataset page: https://huggingface.co/datasets/Starrrrrry/RLAIF-V-Bias-Dataset.