It really surprises me how far we can push a 7B model. It feels like with the right data mix and a 70B range model, we could already be able to match or even out-perform GPT 3.5 with an open-source model!
🔥Open-source, open-science, and data curation for the win!
Meet Notus 7B, a new LLM tuned with DPO on a new curated UltraFeedback dataset, surpassing Zephyr and Claude 2 on AlpacaEval.
Built on the shoulders of giants: 🙌@huggingface Alignment Handbook
argilla.io/blog/notus7b
Dec 1, 2023 · 2:13 PM UTC
2
7
34