Malicious Python package detection on PyPI-based dataset Small Import Amounts, N=1647 (test)
97.29AccuracyH2GLM
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| H2GLM2026.07 | 97.29 | 95.52 | 99.23 | 97.34 | |
| MPHunter2026.07 | 91.15 | 86.1 | 94 | 89.88 | |
| HERCULE2026.07 | 90.22 | 90.97 | 89.31 | 90.13 | |
| Qwen3-14bReasoning-enhanced=true2026.07 | 90.05 | 97.87 | 81.89 | 89.17 | |
| DeepSeek-R1-70bReasoning-enhanced=true2026.07 | 73.85 | 72.86 | 76.02 | 74.41 | |
| Warehouse2026.07 | 70.31 | 92.94 | 43.96 | 59.69 | |
| OSSGadget2026.07 | 66.45 | 56.32 | 93.8 | 70.38 | |
| Naive Bayes2026.07 | 55.8 | 56.1 | 63.5 | 59.57 | |
| Llama3-8bReasoning-enhanced=true2026.07 | 50.64 | 50.42 | 76.91 | 60.91 | |
| Qwen2.5-7bReasoning-enhanced=true2026.07 | 47.13 | 46.69 | 40.43 | 43.34 | |
| Bandit4Mal2026.07 | 46.8 | 41.2 | 23.6 | 30.01 | |
| Clustering2026.07 | 45.12 | 42.54 | 27.8 | 33.63 |