metadata

license: mit
language:
  - en
base_model:
  - Efficient-Large-Model/VILA1.5-13b
pipeline_tag: video-text-to-text

LiFT: Leveraging Human Feedback for Text-to-Video Model Alignment

Summary

This is the model checkpoint proposed in our paper "LiFT: Leveraging Human Feedback for Text-to-Video Model Alignment". LiFT-Critic is a novel Video-Text-to-Text Reward Model for synthesized video evaluation.

Project: https://codegoat24.github.io/LiFT/

Code: https://github.com/CodeGoat24/LiFT

🔧 Installation

Clone the github repository and navigate to LiFT folder

git clone https://github.com/CodeGoat24/LiFT.git
cd LiFT

Install packages

bash ./environment_setup.sh lift

🚀 Inference

Run

Please download this public LiFT-Critic-13b-lora-v1.5 checkpoints.

We provide some synthesized videos for quick inference in ./demo directory.

python LiFT-Critic/test/run_critic_13b.py --model-path ./LiFT-Critic-13b-lora-v1.5

🖊️ Citation

If you find our work helpful, please cite our paper.

@article{LiFT,
  title={LiFT: Leveraging Human Feedback for Text-to-Video Model Alignment.},
  author={Wang, Yibin and Tan, Zhiyu, and Wang, Junyan and Yang, Xiaomeng and Jin, Cheng and Li, Hao},
  journal={arXiv preprint arXiv:2412.04814},
  year={2024}
}