MultiPL-E HumanEval (PHP)
Code · 2022-08-17
The PHP translation of HumanEval from MultiPL-E, which compiles the Python problems' signatures, docstrings and unit tests into other languages, dropping problems that cannot be translated. A problem is solved when the generated PHP code passes the translated tests; scores are pass@1.
Top models (higher is better)
| Model | Score |
|---|---|
| Llama 3.1 405B Instruct | 76.4 |
| Llama 3.1 70B Instruct | 67.7 |
| Code Llama Instruct 70B | 58.4 |
| Llama 3.1 8B Instruct | 54.7 |
| Code Llama Python 70B | 53.4 |
| Code Llama 70B | 50.9 |
| Code Llama 34B | 44.1 |
| Code Llama Python 34B | 42.9 |
| Code Llama Instruct 34B | 36.6 |
| Llama 2 70B | 34.2 |
| Code Llama 13B | 34.2 |
| Code Llama Python 13B | 33.5 |
| Code Llama Instruct 13B | 32.3 |
| Code Llama Python 7B | 32.3 |
| Code Llama Instruct 7B | 28.6 |
| Code Llama 7B | 24.2 |
| Llama 2 34B | 19.9 |
| Llama 2 13B | 13.1 |
| Llama 2 7B | 9.9 |