All credits to
https://huggingface.co/datasets/ai-forever/MERA
This is short version of ai-forever/MERA, where validation (for ['rcb', 'use', 'rwsd', 'parus']) or train (for ['rutie', 'multiq', 'rummlu', 'ruworldtree', 'ruopenbookqa']) was converted to test and random 10 samples from train were converted to prompt split.
Only the following datasets were taken from MERA:
['rcb',
'use',
'rwsd',
'parus',
'rutie',
'multiq',
'rummlu',
'ruworldtree',
'ruopenbookqa']
Raw "instruction" columns… See the full description on the dataset page:
https://huggingface.co/datasets/RefalMachine/darumeru.
