Note A multi-turn evaluation benchmark for chatbots which uses GPT-4 as a judge to evaluate the quality of responses.

Note A high quality blend of human and synthetic datasets focused on dialogues and reasoning abilities. A good source for SFT and PPO.

meta-llama/Llama-2-7b-chat-hf

Text Generation • Updated Apr 17 • 758k • • 3.98k

Note The first series of open access models trained at scale using RLHF. Based on https://huggingface.co/papers/2307.09288

meta-llama/Llama-2-70b-chat-hf

Text Generation • Updated Apr 17 • 181k • 2.16k

Note The first series of open access models trained at scale using RLHF. Based on https://huggingface.co/papers/2307.09288

lightonai/alfred-40b-0723

Text Generation • Updated Aug 11, 2023 • 26 • 45

Note An RLHF tuned version of Falcon 40B

Note A dataset of dialogues between human annotators and a 52B parameter language model from Anthropic. Contains "helpfulness" and "harmlessness" subsets that can be used for training reward models. Basis for https://huggingface.co/papers/2112.00861 and https://huggingface.co/papers/2204.05862