tags | ||
---|---|---|
|
A open-sourced LLM released by Meta and introduced by Touvron et al. (Jul 2023).
The authors react to a fact that:
- there are few open-sourced LLMs "players" due to high computational requirements
- the open-sourced variants of closed product-like LLMs are not as fine-tuned to align with humans and therefore aren't as helpful
- aligning with humans is very costly and currently not very transparent or reproducible