Reported on 24 benchmarks across 8 tasks · 2 papers · 13 SOTA
Note: results are matched by exact model name. Different papers may use the same name for different model variants.