Assembly Code Reverse Engineering on REx86 (test)
5.4166CECodeGemma-7B
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| CodeGemma-7BModel State=Base Model2025.10 | 5.4166 | 16.441 | — | — | |
| CodeLlama-7BModel State=Base Model2025.10 | 2.2446 | 51.34 | — | — | |
| CodeLlama-13BModel State=Base Model2025.10 | 2.1346 | 43.549 | — | — | |
| Qwen2.5-Coder-32BModel State=Base Model2025.10 | 2.0396 | 49.002 | — | — | |
| CodeLlama-34BModel State=Base Model2025.10 | 2.0396 | 45.939 | — | — | |
| Qwen2.5-Coder-3BModel State=Base Model2025.10 | 1.9498 | 39.552 | — | — | |
| Qwen2.5-Coder-14BModel State=Base Model2025.10 | 1.7889 | 56.453 | — | — | |
| Qwen2.5-Coder-7BModel State=Base Model2025.10 | 1.5048 | 55.608 | — | — | |
| Qwen2.5-Coder-3BModel State=Fine-Tuned Model2025.10 | 0.6601 | 49.081 | -66.1 | 24.1 | |
| Qwen2.5-Coder-14BModel State=Fine-Tuned Model2025.10 | 0.6101 | 46.95 | -65.9 | -16.8 | |
| Qwen2.5-Coder-32BModel State=Fine-Tuned Model2025.10 | 0.592 | 55.218 | -71 | 12.7 | |
| CodeGemma-7BModel State=Fine-Tuned Model2025.10 | 0.5448 | 17.235 | -89.9 | 4.8 | |
| REx86Model State=Fine-Tuned Model, Base Model=Qwen2.5-Coder-7B2025.10 | 0.539 | 66.89 | -64.2 | 20.3 | |
| CodeLlama-7BModel State=Fine-Tuned Model2025.10 | 0.4771 | 69.189 | -78.7 | 34.7 | |
| CodeLlama-13BModel State=Fine-Tuned Model2025.10 | 0.4633 | 45.155 | -78.3 | 3.7 | |
| CodeLlama-34BModel State=Fine-Tuned Model2025.10 | 0.4423 | 54.805 | -78.3 | 19.3 |