Llama-2 70B (on 100 first questions, 4-shot, auto-optimized prompting)

Reported on 2 benchmarks across 1 task · 1 paper

Note: results are matched by exact model name. Different papers may use the same name for different model variants.