Code Generation on Coding Race (test)
57Correctness RateGPT-4o
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| GPT-4ocor%_highest_bias_group=Asian, inc%_highest_bias_group=White2025.05 | 57 | 46 | |
| Gemini-Pro-1.5cor%_highest_bias_group=Asian, inc%_highest_bias_group=Black2025.05 | 28 | 42 | |
| LLaMA-3-8bcor%_highest_bias_group=Asian, inc%_highest_bias_group=White2025.05 | 22 | 16 | |
| Claude-3cor%_highest_bias_group=Asian, inc%_highest_bias_group=Black2025.05 | 18 | 28 | |
| GPT-3.5-turbocor%_highest_bias_group=Asian, inc%_highest_bias_group=Black2025.05 | 10 | 10 |