|
--- |
|
task_categories: |
|
- video-text-to-text |
|
license: mit |
|
tags: |
|
- video-generation |
|
- reward-modeling |
|
--- |
|
|
|
This repository contains the implementation of the paper "MJ-VIDEO: Fine-Grained Benchmarking and Rewarding Video Preferences in Video Generation". |
|
|
|
Paper: https://arxiv.org/abs/2502.01719 |
|
Code: https://github.com/aiming-lab/MJ-Video |
|
Project Page: https://aiming-lab.github.io/MJ-VIDEO.github.io/ |
|
|
|
This repository contains the implementation of the paper "MJ-VIDEO: Fine-Grained Benchmarking and Rewarding Video Preferences in Video Generation". We create a fine-grained video preference dataset **MJ-BENCH-VIDEO** and a MoE-structured video reward model **MJ-VIDEO**. |
