Toxicity Detection on Wikipedia Detox
91.3AccuracyGNN-based framework
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| GNN-based framework2025.12 | 91.3 | 91.9 | 90.1 | 91 | 97 | |
| claude-sonnet-4version=42025.12 | 89.9 | 90.6 | 88.6 | 89.6 | 96.3 | |
| nova-pro2025.12 | 89.6 | 88.7 | 90 | 89.3 | 95.4 | |
| llama3.3-70bparameters=70b2025.12 | 89.5 | 85.7 | 94.3 | 89.8 | 94.7 | |
| qwen3-32bparameters=32b2025.12 | 89.5 | 90 | 88.4 | 89.2 | 95.8 | |
| llama4-scout-17bparameters=17b2025.12 | 89.3 | 87.2 | 91.6 | 89.4 | 93.7 | |
| llama4-maverick-17bparameters=17b2025.12 | 89.3 | 86.6 | 92.5 | 89.4 | 95.4 | |
| claude-sonnet-3.7version=3.72025.12 | 89.3 | 85 | 95 | 89.7 | 96 | |
| nova-premier2025.12 | 88.9 | 83.9 | 94.3 | 88.8 | 95.6 | |
| qwen3-235bparameters=235b2025.12 | 88.3 | 83.5 | 94.9 | 88.8 | 94.4 | |
| mistral-large2025.12 | 88.1 | 82.3 | 96.5 | 88.8 | 95.4 | |
| llama3.2-11bparameters=11b2025.12 | 87.5 | 80.9 | 97.2 | 88.3 | 91.7 | |
| nova-lite2025.12 | 83.2 | 92.1 | 71.3 | 80.4 | 84.1 |