Aggression Detection on Wikipedia Detox
90.4AccuracyGNN-based framework
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| GNN-based framework2025.12 | 90.4 | 89.6 | 88.9 | 89.2 | 96.2 | |
| claude-sonnet-3.72025.12 | 88.8 | 86.7 | 88.7 | 87.7 | 95.3 | |
| llama3.3-70bParameters=70b2025.12 | 88 | 83.1 | 91.9 | 87.3 | 93.1 | |
| mistral-large2025.12 | 88 | 83.2 | 91.7 | 87.3 | 94.5 | |
| llama4-scout-17bParameters=17b2025.12 | 87.6 | 83.5 | 90.1 | 86.7 | 93.3 | |
| nova-premier2025.12 | 87.4 | 89 | 79.9 | 84.2 | 94.8 | |
| claude-sonnet-42025.12 | 87.3 | 89.9 | 80.6 | 85 | 95 | |
| nova-pro2025.12 | 87.1 | 86.4 | 84 | 85.2 | 94 | |
| llama4-maverick-17bParameters=17b2025.12 | 86.8 | 82.5 | 89.6 | 85.9 | 92.5 | |
| qwen3-235bParameters=235b2025.12 | 86.8 | 84.2 | 86.8 | 85.5 | 93.5 | |
| qwen3-32bParameters=32b2025.12 | 86.4 | 90.2 | 78.1 | 83.7 | 93.8 | |
| llama3.2-11bParameters=11b2025.12 | 86.1 | 79.6 | 92.6 | 85.6 | 91 | |
| nova-lite2025.12 | 78.2 | 91 | 56.4 | 69.6 | 86.1 |