New base models with Llama 2, pretraining LLMs with ReLoRA, training with less data via AlpaGasus, and many more … There's been lots of interesting research in the last month! I summarized the research highlights in three sentences each:

→ View original post on X — @rasbt