Malicious Python package detection on PyPI-based dataset Large Import Amount, N=487 (test)
0.9584AccuracyH2GLM
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| H2GLM2026.07 | 0.9584 | 0.9514 | 0.9662 | 0.9588 | |
| HERCULE2026.07 | 0.8511 | 0.785 | 0.9671 | 0.8666 | |
| MPHunter2026.07 | 0.7972 | 0.748 | 0.772 | 0.7598 | |
| Warehouse2026.07 | 0.7444 | 0.9161 | 0.538 | 0.6779 | |
| Qwen3-14bReasoning-enhanced=true2026.07 | 0.6783 | 0.9539 | 0.3747 | 0.538 | |
| OSSGadget2026.07 | 0.5492 | 0.442 | 0.84 | 0.5792 | |
| DeepSeek-R1-70bReasoning-enhanced=true2026.07 | 0.4742 | 0.4638 | 0.3307 | 0.3861 | |
| Naive Bayes2026.07 | 0.4538 | 0.4148 | 0.29 | 0.3414 | |
| Bandit4Mal2026.07 | 0.3614 | 0.2816 | 0.165 | 0.2081 | |
| Qwen2.5-7bReasoning-enhanced=true2026.07 | 0.354 | 0.29 | 0.2016 | 0.2378 | |
| Clustering2026.07 | 0.3465 | 0.1505 | 0.0661 | 0.0919 | |
| Llama3-8bReasoning-enhanced=true2026.07 | 0.208 | 0.1753 | 0.1576 | 0.166 |