Reported on 16 benchmarks across 2 tasks · 1 paper · 6 SOTA
Note: results are matched by exact model name. Different papers may use the same name for different model variants.