Meme Classification on HarMeme
88.2AccuracyLLaVA-1.5-7B
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| LLaVA-1.5-7BEvaluation Protocol=RA-HMD2025.02 | 88.2 | 0.935 | |
| Qwen2-VL-7BEvaluation Protocol=RA-HMD2025.02 | 88.1 | 0.932 | |
| Qwen2-VL-2BEvaluation Protocol=RA-HMD2025.02 | 87.7 | 0.929 | |
| RGCLTraining=Supervised fine-tuned2025.02 | 87 | 0.918 | |
| Qwen2-VL-7BEvaluation Protocol=SFT2025.02 | 85.9 | 0.918 | |
| HateCLIPperTraining=Supervised fine-tuned2025.02 | 84.8 | 0.897 | |
| Qwen2-VL-2BEvaluation Protocol=SFT2025.02 | 82.5 | 0.902 | |
| RA-HMDbackbone=LLaVA-1.5-7B, fine-tuning set=HatefulMemes, protocol=RKC2025.02 | 81.9 | 89.5 | |
| RA-HMDbackbone=Qwen2-VL-7B, fine-tuning set=HatefulMemes, protocol=RKC2025.02 | 81.7 | 88.8 | |
| ISSUES2025.02 | 81.6 | 0.928 | |
| LLaVA-1.5-7BEvaluation Protocol=SFT2025.02 | 79.1 | 0.914 | |
| RA-HMDbackbone=Qwen2-VL-2B, fine-tuning set=HatefulMemes, protocol=RKC2025.02 | 78.4 | 86 | |
| CLIPTraining=Supervised fine-tuned2025.02 | 76.8 | 0.826 | |
| LOREHMmode=low-resourced2025.02 | 73.7 | — | |
| GPT-4o2025.02 | 72.9 | — | |
| Mod-Hatemode=low-resourced2025.02 | 69.5 | 73.4 | |
| GPT-4omode=low-resourced2025.02 | 68.4 | — | |
| Qwen2-VL-7BEvaluation Protocol=4-shot2025.02 | 67.2 | 0.715 | |
| Qwen2-VL-2BEvaluation Protocol=4-shot2025.02 | 65.8 | 0.621 | |
| Qwen2-VL-2Bfine-tuning set=HatefulMemes, protocol=SFT + few-shot2025.02 | 65 | 57.4 | |
| Qwen2-VL-7BEvaluation Protocol=zero-shot2025.02 | 64.1 | 0.648 | |
| Qwen2-VL-7Bfine-tuning set=HatefulMemes, protocol=SFT + few-shot2025.02 | 62.4 | 67.2 | |
| RGCLfine-tuning set=HatefulMemes2025.02 | 61.1 | 64.3 | |
| LLaVA-1.5-7BEvaluation Protocol=4-shot2025.02 | 59.6 | 0.734 | |
| Qwen2-VL-2BEvaluation Protocol=zero-shot2025.02 | 56.7 | 0.611 | |
| Qwen2-VL-7Bfine-tuning set=HatefulMemes, protocol=SFT + zero-shot2025.02 | 55.2 | 63 | |
| LLaVA-1.5-7Bfine-tuning set=HatefulMemes, protocol=SFT + few-shot2025.02 | 52.8 | 69.9 | |
| Qwen2-VL-2Bfine-tuning set=HatefulMemes, protocol=SFT + zero-shot2025.02 | 52.2 | 61.3 | |
| LLaVA-1.5-7BEvaluation Protocol=zero-shot2025.02 | 48.6 | 0.714 | |
| LLaVA-1.5-7Bfine-tuning set=HatefulMemes, protocol=SFT + zero-shot2025.02 | 48.5 | 61.9 |