LLaVANext-OmniAlign-32B-DPO

Introduction

Paper: Paper,

Github: Github,

Page: Page,

SFT Dataset: OmniAlign-V,

DPO Dataset: OmniAlign-V-DPO,

MM-AlignBench: VLMEvalkit, Huggingface

Checkpoints: LLaVANext-OA-7B, LLaVANext-OA-32B, LLaVANext-OA-32B-DPO

This is the official repo of LLaVANext-OmniAlign(OA)-32B-DPO in OmniAlign-V: Towards Enhanced Alignment of MLLMs with Human Preference.

LLaVANext-OmniAlign-32B-DPO is based on LLaVA-Next structure with Qwen2.5-32B-Instruct.

By applying DPO stage using OmniAlign-V-DPO datasets, we can further improve the alignment of MLLMs with human preference.

Performance

By integrating OmniAlign-V-DPO datasets in DPO stage, we can further improve the alignment of MLLMs with human preference. Our LLaVANext-OA-32B-DPO even surpasses Qwen2VL-72B on MM-AlignBench.

How to use

Please refer to our Github for more details about training and evaluation.