top | item 38581855 (no title) cicce19 | 2 years ago Does anyone know the advantage of mistral over llama2? I understand the founders of mistral previously worked on llama. discuss order hn newest vinni2|2 years ago Mistral 7b version is slightly better than 7b llama based on elo rating https://huggingface.co/spaces/lmsys/chatbot-arena-leaderboar... MacsHeadroom|2 years ago It's better than that. The best Llama-2 finetune is 88th on the 7B leaderboard, behind 87 Mistral finetunes.Mistral 7B is above every Llama-2 finetune, including the 10x larger llama-2 70B, on the Open LLM Leaderboard.https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderb...(Note: Check "architecture" and uncheck precision "?" then ignore "tigerbot" as it was disqualified for cheating.) nuz|2 years ago Also less censored load replies (1)
vinni2|2 years ago Mistral 7b version is slightly better than 7b llama based on elo rating https://huggingface.co/spaces/lmsys/chatbot-arena-leaderboar... MacsHeadroom|2 years ago It's better than that. The best Llama-2 finetune is 88th on the 7B leaderboard, behind 87 Mistral finetunes.Mistral 7B is above every Llama-2 finetune, including the 10x larger llama-2 70B, on the Open LLM Leaderboard.https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderb...(Note: Check "architecture" and uncheck precision "?" then ignore "tigerbot" as it was disqualified for cheating.) nuz|2 years ago Also less censored load replies (1)
MacsHeadroom|2 years ago It's better than that. The best Llama-2 finetune is 88th on the 7B leaderboard, behind 87 Mistral finetunes.Mistral 7B is above every Llama-2 finetune, including the 10x larger llama-2 70B, on the Open LLM Leaderboard.https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderb...(Note: Check "architecture" and uncheck precision "?" then ignore "tigerbot" as it was disqualified for cheating.)
vinni2|2 years ago
MacsHeadroom|2 years ago
Mistral 7B is above every Llama-2 finetune, including the 10x larger llama-2 70B, on the Open LLM Leaderboard.
https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderb...
(Note: Check "architecture" and uncheck precision "?" then ignore "tigerbot" as it was disqualified for cheating.)
nuz|2 years ago