Fine-grained Multi-label Threat Categorization on GUARDZOO (out-of-domain)
0.3101Macro F1ROUTEGUARD-4B
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| ROUTEGUARD-4BImplementation Strategy=Router (Proposed), Evaluation Setting=Out-of-domain2026.05 | 0.3101 | 0.5308 | |
| ROUTEGUARD-8BImplementation Strategy=Router (Proposed), Evaluation Setting=Out-of-domain2026.05 | 0.3037 | 0.53 | |
| LLAMAGUARD3-8BImplementation Strategy=Fine-tuned Baselines, Evaluation Setting=Out-of-domain2026.05 | 0.2893 | 0.4638 | |
| MLP ROUTERImplementation Strategy=Router Baselines, Evaluation Setting=Out-of-domain2026.05 | 0.2623 | 0.3779 | |
| SHIELDGEMMA-9BImplementation Strategy=Fine-tuned Baselines, Evaluation Setting=Out-of-domain2026.05 | 0.2452 | 0.4425 | |
| GRAPH ROUTERImplementation Strategy=Router Baselines, Evaluation Setting=Out-of-domain2026.05 | 0.2384 | 0.3878 | |
| DC ROUTERImplementation Strategy=Router Baselines, Evaluation Setting=Out-of-domain2026.05 | 0.2326 | 0.3781 | |
| QWEN3GUARD-4BImplementation Strategy=Fine-tuned Baselines, Evaluation Setting=Out-of-domain2026.05 | 0.2055 | 0.3459 | |
| QWEN3GUARD-8BImplementation Strategy=Fine-tuned Baselines, Evaluation Setting=Out-of-domain2026.05 | 0.1923 | 0.2308 | |
| KNN ROUTERImplementation Strategy=Router Baselines, Evaluation Setting=Out-of-domain2026.05 | 0.1697 | 0.1934 |