1 paper · 1 filter
Khalil Hennara, Ahmad Bastati, Muhammad Hreden +4
The performance of large language models (LLMs) and large multimodal models (LMMs) depends heavily on the quality and scale of their pre-training datasets. Recent research shows th…