HLE (Humanity's Last Exam)
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
HLE (Humanity's Last Exam) text-only subset (val)
52.2Inference Accuracy
13
HLE (Humanity's Last Exam) 2% subset
2.9MAE
3
HLE (Humanity's Last Exam) 1% subset
3.5MAE
3
HLE (Humanity's Last Exam) 0.5% subset
5.6MAE
3