Reported on 9 benchmarks across 4 tasks · 4 papers · 3 SOTA
Note: results are matched by exact model name. Different papers may use the same name for different model variants.