@article{mugenaplaygroundforvideoaudiotext,
  title = {MUGEN: A Playground for Video-Audio-Text Multimodal Understanding and GENeration},
  author = {Thomas Hayes and Songyang Zhang and Xi Yin and Guan Pang and Sasha Sheng and Harry Yang and Songwei Ge and Qiyuan Hu and Devi Parikh},
  year = {2022},
  eprint = {2204.08058},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2204.08058v3},
}