bigscience-workshop/evaluation

Add DiaBLa to Full Benchmark

epavlick opened this issue · 5 comments

Add DiaBLa to Full Benchmark

I can do this!

looks like this needs to be added to HF first

Hi, sorry - I got very busy! I will try to have this added by the end of tomorrow!

@rbawden will this be ready? we can add it to HF otherwise

Hi - yes, it's added here: https://huggingface.co/datasets/rbawden/DiaBLa. I am currently creating some prompts for various tasks, including (i) basic sentence-level translation, (ii) contextual translation (same language previous sentence and original language previous sentence) and (iii) distinguishing MT output from reference output