MTSAIR

/

multi_verse_model

like 31

@Kquant03 Thanks for sharing your document very interesting thinking, the details about our approach is different hopefully will be published soon and would love to share, stay tuned for new models as well

@Kquant03 Thanks for sharing your document very interesting thinking, the details about our approach is different hopefully will be published soon and would love to share, stay tuned for new models as well

Yeah, we have a couple of different projects cooking too. I'm super excited to read what you guys write...Most people recommend LaTeX for papers.

Hey there @ammarali32 and @Kquant03 !

I'm working on a project that uses an Open Source LLM as the base for a tool to process large books chapter-wise (e.g., summarize 500-page books into concise chapters). I came across the Multiversal model and am interested in its suitability for this task. The idea is to utilize the multiversal concept to generate various summary options for each chapter, then use a separate critiquing model to assess their quality using the original context.

Please share your thoughts on whether this use case aligns with the model's capabilities and if you have any recommendations for other models that might better suit this purpose. I also have difficulty finding suitable LLMs specialized for different tasks (like 1 for summarization and another for critiquing for example) on the LLM leaderboards. If you had any insight in how to find specialized models I would appreciate your expertise so much!

Thank you!

Hi @mruniverse009 ,interesting project, I don't think it will support your use case without further tuning/training, i would suggest you to check the following:
https://arxiv.org/html/2401.03462v2
https://github.com/dvlab-research/LongLoRA/tree/main?tab=readme-ov-file
There are multiple research with that but these are the fresh recent studies as i expect