openbmb

/

MiniCPM-Llama3-V-2_5

like 1.3k

I am a Korean developer, and I am interested in Korean benchmarks.

While I was surprised by the model's Korean score compared with other models, I got two curious question

1. I couldn't find the same score on other models' card. So did you get that score by your own test?

2. LLAVA Benchmark actually do not provide Korean dataset. So did I wonder how you made it and its step by step details

Also thank you for releasing such a fancy model~