Out-of-Domain Reasoning Suite
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
Out-of-Domain Reasoning Suite
94.5ARC-c Score
29
Out-of-Domain (OOD) Reasoning Suite (ARC-c, GPQA*, MMLU-Pro) various (test)
93.9ARC-c Accuracy (avg@4)
21
Out-of-Domain Reasoning Suite BGQA, CRUX Eval, Strategy QA, Table Bench
71.1BGQA Accuracy
9