马会东的博客
先说结论：截至目前，在开源大模型阵营中Meta Llama3 70B在Code Generation方面的表现是最棒的
from https://ai.meta.com/blog/meta-llama-3/
from https://github.com/jamesmurdza/humaneval-results
from https://klu.ai/glossary/humaneval-benchmark
