README.md

---
license: apache-2.0
---

# VCoder LLaVA-1.5-13b

VCoder LLaVA-1.5-13b was trained on COST training dataset in December 2023. It uses the pretrained [LLaVA-1.5-13b](https://huggingface.co/liuhaotian/llava-v1.5-13b) model weights. It was introduced by Jain et al. in [this repository](https://github.com/SHI-Labs/VCoder).

VCoder is an adapter for improving existing Multimodal LLMs at object-level perception tasks with the use of perception modalities as control inputs while retaining performance on other tasks.

![img](https://praeclarumjj3.github.io/vcoder/vcoder.svg)

### Citation

```bibtex
@article{jain2023vcoder,
    title={{VCoder: Versatile Vision Encoders for Multimodal Large Language Models}},
    author={Jitesh Jain and Jianwei Yang and Humphrey Shi},
    journal={arXiv},
    year={2023}
}
```