I generate code by gpt-4 with prompt in dataset, and evaluate the code in `codegeex/codegeex:0.1.23`. the result is `'pass@1': 47.1` that cannot match `'pass@1': 64.3` in the paper. <img width="1821" alt="image" src="https://github.com/THUDM/NaturalCodeBench/assets/8615337/d304f4fd-cabc-4a4e-8a79-560cde219bd9"> the gpt-4 output is here [gpt4_ncb_java_zh.jsonl.zip](https://github.com/user-attachments/files/15884262/gpt4_ncb_java_zh.jsonl.zip)
I generate code by gpt-4 with prompt in dataset, and evaluate the code in
codegeex/codegeex:0.1.23.the result is
'pass@1': 47.1that cannot match'pass@1': 64.3in the paper.the gpt-4 output is here
gpt4_ncb_java_zh.jsonl.zip