MultiPL-E HumanEval (Java)
Code · 2022-08-17
The Java 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 Java code passes the translated tests; scores are pass@1.
Top models (higher is better)
| Model | Score |
|---|---|
| Llama 3.1 405B Instruct | 80.4 |
| Llama 3.1 70B Instruct | 72.2 |
| Code Llama Instruct 70B | 58.2 |
| Llama 3.1 8B Instruct | 58.2 |
| Code Llama Python 70B | 57.6 |
| Code Llama 70B | 51.9 |
| Code Llama 34B | 45.6 |
| Code Llama Python 34B | 44.9 |
| Code Llama Instruct 34B | 43.7 |
| Code Llama Instruct 13B | 40.5 |
| Code Llama 13B | 38.0 |
| Code Llama Python 13B | 37.3 |
| Code Llama Python 7B | 35.4 |
| Code Llama 7B | 34.2 |
| Llama 2 70B | 31.7 |
| Code Llama Instruct 7B | 30.4 |
| Llama 2 34B | 22.2 |
| Llama 2 13B | 15.8 |
| Llama 2 7B | 10.8 |