Meme Classification on PrideMM
79.2AccuracyMemeGuard
Evaluation results
| Method | Links | ||
|---|---|---|---|
| MemeGuardTraining Strategy=Three-stage training, Backbone=Qwen2.5-VL-7B-Instruct2025.06 | 79.2 | 78.9 | |
| LLaVA-1.5-7BEvaluation Protocol=RA-HMD2025.02 | 78.1 | 78.7 | |
| Qwen2-VL-7BEvaluation Protocol=RA-HMD2025.02 | 78.1 | 78.4 | |
| SFT on MemeMindTraining Strategy=Supervised Fine-Tuning, Training Dataset=MemeMind2025.06 | 78 | 77.8 | |
| RGCLTraining=Supervised fine-tuned2025.02 | 76.3 | 76.5 | |
| MemeCLIP2025.02 | 76.1 | 75.1 | |
| Qwen2-VL-2BEvaluation Protocol=RA-HMD2025.02 | 76 | 76.7 | |
| HateCLIPperTraining=Supervised fine-tuned2025.02 | 75.5 | 74.1 | |
| GPT-4o2025.02 | 75.3 | 73.7 | |
| Qwen2-VL-7BEvaluation Protocol=SFT2025.02 | 75.1 | 74.9 | |
| Qwen2-VL-2BEvaluation Protocol=SFT2025.02 | 73.7 | 74.2 | |
| LLaVA-1.5-7BEvaluation Protocol=SFT2025.02 | 73.2 | 76 | |
| CLIPTraining=Supervised fine-tuned2025.02 | 72.4 | 72.3 | |
| RA-HMDbackbone=Qwen2-VL-7B, fine-tuning set=HatefulMemes, protocol=RKC2025.02 | 69.3 | 69.3 | |
| Qwen2-VL-7BEvaluation Protocol=4-shot2025.02 | 69.1 | 56.6 | |
| RA-HMDbackbone=LLaVA-1.5-7B, fine-tuning set=HatefulMemes, protocol=RKC2025.02 | 68.8 | 67.7 | |
| Qwen2.5-VL-7B-Instructmode=Zero-shot2025.06 | 66.8 | 64.1 | |
| Qwen2-VL-7BEvaluation Protocol=zero-shot2025.02 | 65.3 | 62.9 | |
| Qwen2-VL-7Bfine-tuning set=HatefulMemes, protocol=SFT + zero-shot2025.02 | 64.5 | 63.6 | |
| GPT-4omode=low-resourced2025.02 | 63.8 | 62.3 | |
| LLaVA-1.5-7BEvaluation Protocol=zero-shot2025.02 | 63.4 | 65.6 | |
| LLaVA-1.5-7BEvaluation Protocol=4-shot2025.02 | 62.1 | 64 | |
| RGCLfine-tuning set=HatefulMemes2025.02 | 59.8 | 61.5 | |
| LLaVA-1.5-7Bfine-tuning set=HatefulMemes, protocol=SFT + zero-shot2025.02 | 58.1 | 53.3 | |
| Qwen2-VL-2BEvaluation Protocol=zero-shot2025.02 | 57.8 | 53.3 | |
| Qwen2-VL-2Bfine-tuning set=HatefulMemes, protocol=SFT + zero-shot2025.02 | 56.4 | 59 | |
| Qwen2-VL-2Bfine-tuning set=HatefulMemes, protocol=SFT + few-shot2025.02 | 55.4 | 60.9 | |
| Qwen2-VL-7Bfine-tuning set=HatefulMemes, protocol=SFT + few-shot2025.02 | 55.4 | 60.9 | |
| Qwen2-VL-2BEvaluation Protocol=4-shot2025.02 | 55.4 | 54.3 | |
| RA-HMDbackbone=Qwen2-VL-2B, fine-tuning set=HatefulMemes, protocol=RKC2025.02 | 53.4 | 69 | |
| LLaVA-1.5-7Bfine-tuning set=HatefulMemes, protocol=SFT + few-shot2025.02 | 48.5 | 55.3 |