Excited to release Tulu 3! We worked hard to try and make the best open post-training recipe we could, and the results are good!
I was lucky enough to work on almost every stage of the pipeline in one way or another. Some comments + highlights ⬇️
I was lucky enough to work on almost every stage of the pipeline in one way or another. Some comments + highlights ⬇️
Comments
8B model: https://buff.ly/498x15q
70B model: https://buff.ly/3Ok4PTp
Demo: https://buff.ly/492H2Rw
Website: https://allenai.org/tulu
Working out the best ways to generate synthetic data was crucial to really boosting performance.
The gains are smaller when your base models are already strong, but I am excited to take this further!