WikiBigEdit
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
WikiBigEdit
81.22Efficacy
63
WikiBigEdit
79.6Efficacy
49
WikiBigEdit
69.5MMLU
34
WikiBigEdit
1Error Rate (ERR)
24
WikiBigEdit 3,000 samples (test)
99.9Reliability
13
WikiBigEdit (500K edits)
74.54Efficacy
12
WikiBigEdit 500K edits
74.54Efficacy
9
WikiBigEdit
99.2Reliability
6