Fine-grained Hate Categorization on H-VLI
86.92AccuracyGPT-5
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| GPT-5Training=False, Framework=ARCADE2026.03 | 86.92 | 81.73 | 86.99 | |
| GPT-5-miniTraining=False, Framework=ARCADE2026.03 | 82.59 | 80.22 | 83.08 | |
| GPT-4oTraining=False, Framework=ARCADE2026.03 | 82.51 | 79.53 | 82.93 | |
| GPT-4.1-miniTraining=False, Framework=ARCADE2026.03 | 80.72 | 74.8 | 80.41 | |
| GPT-5Training=False2026.03 | 80.48 | 74.94 | 81.39 | |
| Gemini-2.5-ProTraining=False, Framework=ARCADE2026.03 | 80.03 | 76.03 | 80.49 | |
| Qwen3-VL-PlusTraining=False, Framework=ARCADE2026.03 | 79.95 | 75.04 | 79.95 | |
| Gemini-2.5-FlashTraining=False, Framework=ARCADE2026.03 | 79.78 | 75.19 | 80.24 | |
| Gemini-2.5-FlashTraining=False2026.03 | 78.61 | 75.33 | 79.26 | |
| Gemini-2.5-ProTraining=False2026.03 | 77.93 | 77.32 | 78.55 | |
| Qwen-VL-MaxTraining=False, Framework=ARCADE2026.03 | 75 | 70.27 | 75.42 | |
| GPT-4oTraining=False2026.03 | 74.36 | 73.19 | 74.94 | |
| GPT-4.1-miniTraining=False2026.03 | 72.24 | 72.26 | 72.92 | |
| Qwen3-VL-PlusTraining=False2026.03 | 70.61 | 69.42 | 71.16 | |
| Qwen-VL-PlusTraining=False, Framework=ARCADE2026.03 | 70.29 | 63.89 | 70.23 | |
| GPT-5-miniTraining=False2026.03 | 70.12 | 68.62 | 70.89 | |
| MMBTTraining=True2026.03 | 66.38 | 62.17 | 66.99 | |
| BERT+ViTTraining=True2026.03 | 65.62 | 61.25 | 66.19 | |
| PromptHateTraining=True2026.03 | 65.03 | 60.55 | 65.47 | |
| Qwen-VL-MaxTraining=False2026.03 | 64.91 | 66.6 | 65.31 | |
| BERTTraining=True2026.03 | 62.31 | 57.4 | 63.03 | |
| Qwen-VL-PlusTraining=False2026.03 | 61.41 | 59.75 | 63.16 | |
| MomentaTraining=True2026.03 | 60.19 | 51.86 | 60.35 | |
| ViT-b-16Training=True2026.03 | 47.88 | 14.52 | 43.32 |