metadata

license: mit
base_model:
  - VITA-MLLM/VITA

MAVEN

We provide LoRA fine-tuned model weights and configurations for MAVEN. These files can be used to reproduce our experimental results.

Resources

Github: https://github.com/reml-group/fortisavqa
Arxiv: https://arxiv.org/abs/2504.00487

Citation

If you use this model in your research, please cite the following paper:

@article{ma2025fortisavqa,
  title={FortisAVQA and MAVEN: a Benchmark Dataset and Debiasing Framework for Robust Multimodal Reasoning},
  author={Ma, Jie and Gao, Zhitao and Chai, Qi and Liu, Jun and Wang, Pinghui and Tao, Jing and Su, Zhou},
  journal={arXiv preprint arXiv:2504.00487},
  year={2025}
}
@inproceedings{malook,
  title={Look, Listen, and Answer: Overcoming Biases for Audio-Visual Question Answering},
  author={Ma, Jie and Hu, Min and Wang, Pinghui and Sun, Wangchun and Song, Lingyun and Pei, Hongbin and Liu, Jun and Du, Youtian},
  booktitle={NeurIPS},
  year={2024}
}