Sealbook

evalplus/evalplus

Rigourous evaluation of LLM-synthesized code - NeurIPS 2023 & COLM 2024

benchmarkchatgptefficiencygpt-4large-language-modelsprogram-synthesistesting