Reasoning and Knowledge Suite
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
Reasoning & Knowledge Suite (ARC-E, WG, SIQA, Hella., OBQA, CSQA, BA, MMLU)
51.05ARC-Easy Accuracy
15
Reasoning and Knowledge Suite (MMLU, ARC-C, ARC-E, BoolQ, CSQA, HSwag, PIQA, SocIQ, Wino) (various)
75.78MMLU
14