ai-forever/MERA Issues
Как добавить форматирование промпта?
Updated 20.4.0 lm-evaluation-harness
Updated 9Скоринг GGUF моделей
Closed 1Значения логов бенчмарка
Updated 1empty value rummlu
Closed 3
MERA (Multimodal Evaluation for Russian-language Architectures) is a new open benchmark for the Russian language for evaluating fundamental models.