Explanation Generation on Ex-ToxiCN-MM
3.14InformativenessEx-ToxiCN-MM (human)
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| Ex-ToxiCN-MM (human)Module=Ex-ToxiCN-MM, Method=human2026.05 | 3.14 | 4.03 | 2.9 | 4.79 | 3.46 | — | — | |
| Qwen-VL-Max (RIKE)Module=Qwen-VL-Max, Method=RIKE2026.05 | 3.11 | 3.9 | 3.13 | 4.45 | 3.39 | 10.17 | 1.01 | |
| GLM-4V-Flash (RIKE)Module=GLM-4V-Flash, Method=RIKE2026.05 | 3.08 | 3.98 | 2.87 | 4.41 | 3.37 | 9.48 | 32.3 | |
| Qwen-VL-Max (Base)Module=Qwen-VL-Max, Method=Base2026.05 | 3.05 | 3.83 | 3.06 | 4.34 | 3.31 | 9.94 | 1.09 | |
| GLM-4V-Flash (Base)Module=GLM-4V-Flash, Method=Base2026.05 | 3.04 | 3.88 | 3.01 | 3.81 | 3.26 | 2.94 | 18.98 | |
| InternVL3.5-8B (RIKE)Module=InternVL3.5-8B, Method=RIKE2026.05 | 2.9 | 3.83 | 2.71 | 4.63 | 3.22 | 53.48 | 64.38 | |
| Qwen2.5-VL-7B (RIKE)Module=Qwen2.5-VL-7B, Method=RIKE2026.05 | 2.84 | 3.76 | 2.64 | 4.57 | 3.15 | 51.69 | 62.51 | |
| InternVL3.5-8B (Base)Module=InternVL3.5-8B, Method=Base2026.05 | 2.75 | 3.7 | 2.63 | 4.79 | 3.14 | 8.86 | 28.28 | |
| Qwen2.5-VL-7B (Base)Module=Qwen2.5-VL-7B, Method=Base2026.05 | 2.73 | 3.64 | 2.51 | 4.5 | 2.97 | 14.17 | 29.31 | |
| LLaVa-1.5-7b (RIKE)Module=LLaVa-1.5-7b, Method=RIKE2026.05 | 2.57 | 3.48 | 2.36 | 4.44 | 2.75 | 51.2 | 44.25 | |
| LLaVa-1.5-7b (Base)Module=LLaVa-1.5-7b, Method=Base2026.05 | 2.52 | 3.32 | 2.58 | 3.77 | 2.64 | 11.2 | 25.64 |