Datasets:

Modalities:
Tabular
Text
Formats:
parquet
Languages:
English
ArXiv:
DOI:
Libraries:
Datasets
Dask
License:
guipenedo HF staff commited on
Commit
a181cbc
1 Parent(s): 9224423

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +1 -2
README.md CHANGED
@@ -660,13 +660,12 @@ To conduct the ablations for each of our dataset filtering choices, we selected
660
 
661
  We used the following list of benchmark for our ablation runs:
662
 
663
- - commonsense_qa (acc_norm)
664
  - hellaswag (acc/acc_norm)
665
  - openbookqa (acc/acc_norm)
666
  - piqa (acc/acc_norm)
667
  - siqa (acc/acc_norm)
668
  - winogrande (acc/acc_norm)
669
- - sciq (acc/acc_norm)
670
  - arc (acc/acc_norm)
671
  - mmlu (acc/acc_norm)
672
 
 
660
 
661
  We used the following list of benchmark for our ablation runs:
662
 
663
+ - commonsense_qa (acc/acc_norm)
664
  - hellaswag (acc/acc_norm)
665
  - openbookqa (acc/acc_norm)
666
  - piqa (acc/acc_norm)
667
  - siqa (acc/acc_norm)
668
  - winogrande (acc/acc_norm)
 
669
  - arc (acc/acc_norm)
670
  - mmlu (acc/acc_norm)
671