Adapting Multimodal Large Language Models to Domains via Post-Training

This repository provides an implementation preview of our paper: On Domain-Specific Post-Training for Multimodal Large Language Models.

We investigate domain adaptation of MLLMs through post-training, focusing on data synthesis, training pipelines, and task evaluation. (1) Data Synthesis: Using open-source models, we develop a visual instruction synthesizer that effectively generates diverse visual instruction tasks from domain-specific image-caption pairs. Our synthetic tasks surpass those generated by manual rules, GPT-4, and GPT-4V in enhancing the domain-specific performance of MLLMs. (2) Training Pipeline: While the two-stage training--initially on image-caption pairs followed by visual instruction tasks--is commonly adopted for developing general MLLMs, we apply a single-stage training pipeline to enhance task diversity for domain-specific post-training. (3) Task Evaluation: We conduct experiments in two domains, biomedicine and food, by post-training MLLMs of different sources and scales (e.g., Qwen2-VL-2B, LLaVA-v1.6-8B, Llama-3.2-11B), and then evaluating MLLM performance on various domain-specific tasks.

***************** Updates ********************

[2024/12/05-11] Released all our data and models
[2024/11/29] Released our paper

Resources

🤗 We share our data and models with example usages, feel free to open any issues or discussions! 🤗

Model	Repo ID in HF 🤗	Domain	Base Model	Training Data	Evaluation Benchmark
Visual Instruction Synthesizer	AdaptLLM/visual-instruction-synthesizer	-	open-llava-next-llama3-8b	VisionFLAN and ALLaVA	-
AdaMLLM-med-2B	AdaptLLM/biomed-Qwen2-VL-2B-Instruct	Biomedicine	Qwen2-VL-2B-Instruct	biomed-visual-instructions	biomed-VQA-benchmark
AdaMLLM-food-2B	AdaptLLM/food-Qwen2-VL-2B-Instruct	Food	Qwen2-VL-2B-Instruct	food-visual-instructions	food-VQA-benchmark
AdaMLLM-med-8B	AdaptLLM/biomed-LLaVA-NeXT-Llama3-8B	Biomedicine	open-llava-next-llama3-8b	biomed-visual-instructions	biomed-VQA-benchmark
AdaMLLM-food-8B	AdaptLLM/food-LLaVA-NeXT-Llama3-8B	Food	open-llava-next-llama3-8b	food-visual-instructions	food-VQA-benchmark
AdaMLLM-med-11B	AdaptLLM/biomed-Llama-3.2-11B-Vision-Instruct	Biomedicine	Llama-3.2-11B-Vision-Instruct	biomed-visual-instructions	biomed-VQA-benchmark
AdaMLLM-food-11B	AdaptLLM/food-Llama-3.2-11B-Vision-Instruct	Food	Llama-3.2-11B-Vision-Instruct	food-visual-instructions	food-VQA-benchmark

Code: https://github.com/bigai-ai/QA-Synthesizer

About

AdaMLLM is our latest effort to enhance task generalization of (M)LLMs by scaling synthetic supervised tasks based on unsupervised contexts.

AdaptLLM
We employ rule-based methods to extract tasks from domain-specific corpora, reformatting them into reading comprehension tasks for continued pre-training. Our 7B finance model outperforms domain-specific models of much larger scales, such as BloombergGPT-50B.
Instruction Pre-Training
We develop a general-purpose instruction synthesizer which significantly increases task diversity for LM pre-training, outperforming vanilla pre-training in both general pre-training from scratch and domain-adaptive continual pre-training.
AdaMLLM
We extend supervised task synthesis to multimodality, introducing a unified visual instruction synthesizer to extract instruction-response pairs from image-caption data. Our synthetic tasks outperform those generated by manual rules, GPT-4, and GPT-4V in improving domain-specific performance for MLLMs.

Looking ahead, we envision further broadening the scope of supervised task synthesis, efficiently enhancing the general capabilities of trained models.

Contact

Daixuan Cheng: daixuancheng6@gmail.com

Citation

If you find our work helpful, please cite us.

Adapt MLLM to Domains

@article{adamllm,
  title={On Domain-Specific Post-Training for Multimodal Large Language Models},
  author={Cheng, Daixuan and Huang, Shaohan and Zhu, Ziyu and Zhang, Xintong and Zhao, Wayne Xin and Luan, Zhongzhi and Dai, Bo and Zhang, Zhenliang},
  journal={arXiv preprint arXiv:2411.19930},
  year={2024}
}

Instruction Pre-Training (EMNLP 2024)

@article{instructPT,
  title={Instruction Pre-Training: Language Models are Supervised Multitask Learners},
  author={Cheng, Daixuan and Gu, Yuxian and Huang, Shaohan and Bi, Junyu and Huang, Minlie and Wei, Furu},
  journal={arXiv preprint arXiv:2406.14491},
  year={2024}
}

Adapt LLM to Domains (ICLR 2024)

@inproceedings{
adaptllm,
title={Adapting Large Language Models via Reading Comprehension},
author={Daixuan Cheng and Shaohan Huang and Furu Wei},
booktitle={The Twelfth International Conference on Learning Representations},
year={2024},
url={https://openreview.net/forum?id=y886UXPEZ0}
}