Hi, I just want to share with all of you two very interesting news in the NLP world:
1) Meta has just released the Llama 4 ecosystem of models, right now only two of them are available Llama 4 Scout and Llama 4 Maverick, both with 17B parameters but with some differences. You con read all the details in their web page here https://ai.meta.com/blog/llama-4-multimodal-intelligence/ .
2) I found this paper by Meta that I think is interesting if someone wants to read [2504.00927] Multi-Token Attention . It is about a new technique developed by Meta called Multi-Token Attention and it has been published recently.
Thanks for your attention
1) Meta has just released the Llama 4 ecosystem of models, right now only two of them are available Llama 4 Scout and Llama 4 Maverick, both with 17B parameters but with some differences. You con read all the details in their web page here https://ai.meta.com/blog/llama-4-multimodal-intelligence/ .
2) I found this paper by Meta that I think is interesting if someone wants to read [2504.00927] Multi-Token Attention . It is about a new technique developed by Meta called Multi-Token Attention and it has been published recently.
Thanks for your attention
Pietro Volpato