Closed treasuraid closed 11 months ago
Thank you for your interest in KoLLaVA.
I have been exploring various evaluation methods for the model. However, I don't have immediate plans to translate and test the ScienceQA dataset in Korean.
If its original meaning remains intact through translation, testing it could be a good direction 👍
First, thank you for great work on KoLLAVA.
how about adding KoScienceQA dataset for instruct tuning? The performance would be competitive or not?