Update README.md

README.md

@@ -6,9 +6,6 @@ language:
 
 This repository provides an implementation preview of our paper: [On Domain-Specific Post-Training for Multimodal Large Language Models](https://huggingface.co/papers/2411.19930).
 
-Our code will be available at [https://github.com/bigai-ai/QA-Synthesizer](https://github.com/bigai-ai/QA-Synthesizer)
-
-
 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. 
@@ -25,12 +22,7 @@ We investigate domain adaptation of MLLMs through post-training, focusing on dat
 
 
 ***************** **Updates** ********************  
-- [2024/12/11] Released [food visual instructions](https://huggingface.co/datasets/AdaptLLM/food-visual-instructions) for post-training MLLMs
-- [2024/12/10] Released evaluation benchmark datasets for biomedicine and food domains: [biomed-VQA-benchmark](https://huggingface.co/datasets/AdaptLLM/biomed-VQA-benchmark), [food-VQA-benchmark](https://huggingface.co/datasets/AdaptLLM/food-VQA-benchmark).
-- [2024/12/9] Released AdaMLLM developed from llava-next-llama3-8b: [AdaMLLM-med-8B](https://huggingface.co/AdaptLLM/biomed-LLaVA-NeXT-Llama3-8B), [AdaMLLM-food-8B](https://huggingface.co/AdaptLLM/food-LLaVA-NeXT-Llama3-8B).
-- [2024/12/7] Released [visual-instruction-synthesizer](https://huggingface.co/AdaptLLM/visual-instruction-synthesizer) used to synthesize task triplets based on image-caption pairs.
-- [2024/12/6] Released AdaMLLM developed from Qwen2-VL-2B and Llama-3.2-11B-Vision: [AdaMLLM-med-2B](https://huggingface.co/AdaptLLM/biomed-Qwen2-VL-2B-Instruct), [AdaMLLM-food-2B](https://huggingface.co/AdaptLLM/food-Qwen2-VL-2B-Instruct), [AdaMLLM-med-11B](https://huggingface.co/AdaptLLM/biomed-Llama-3.2-11B-Vision-Instruct),  [AdaMLLM-food-11B](https://huggingface.co/AdaptLLM/food-Llama-3.2-11B-Vision-Instruct), 
-- [2024/12/05] Released [biomedicine visual instructions](https://huggingface.co/datasets/AdaptLLM/biomed-visual-instructions) for post-training MLLMs
+- [2024/12/05-11] Released all our data and models
 - [2024/11/29] Released our paper
 
 
@@ -47,9 +39,7 @@ We investigate domain adaptation of MLLMs through post-training, focusing on dat
 | [AdaMLLM-med-11B](https://huggingface.co/AdaptLLM/biomed-Llama-3.2-11B-Vision-Instruct) | AdaptLLM/biomed-Llama-3.2-11B-Vision-Instruct     | Biomedicine  | Llama-3.2-11B-Vision-Instruct    | [biomed-visual-instructions](https://huggingface.co/datasets/AdaptLLM/biomed-visual-instructions) | [biomed-VQA-benchmark](https://huggingface.co/datasets/AdaptLLM/biomed-VQA-benchmark)                   |
 | [AdaMLLM-food-11B](https://huggingface.co/AdaptLLM/food-Llama-3.2-11B-Vision-Instruct) | AdaptLLM/food-Llama-3.2-11B-Vision-Instruct     | Food | Llama-3.2-11B-Vision-Instruct    | [food-visual-instructions](https://huggingface.co/datasets/AdaptLLM/food-visual-instructions) |  [food-VQA-benchmark](https://huggingface.co/datasets/AdaptLLM/food-VQA-benchmark)                   |
 
-
-## Contact  
-Daixuan Cheng: `[email protected]`
+Code: [https://github.com/bigai-ai/QA-Synthesizer](https://github.com/bigai-ai/QA-Synthesizer)
 
 ## About
 
@@ -67,6 +57,9 @@ AdaMLLM represents our latest advancement in building domain-specific foundation
   We extend supervised task synthesis to multimodality, introducing a unified visual instruction synthesizer to extract instruction-response pairs from domain-specific image-caption pairs. Our synthetic tasks outperform those generated by manual rules, GPT-4, and GPT-4V in improving domain-specific performance for MLLMs.
 
 
+## Contact  
+Daixuan Cheng: `[email protected]`
+
 
 ## Citation
 If you find our work helpful, please cite us.