Meme Classification 벤치마크
Meme Classification on MultiOFF
Accuracy
- 2021-02-26 — CLIP: Accuracy 62.4
- 2023-11-14 — RGCL: Accuracy 67.1
- 2025-02-18 — RA-HMD (Qwen2-VL-7B): Accuracy 71.1
| Rank | Model | Accuracy | F1 | Paper | Code | Year |
|---|---|---|---|---|---|---|
| 1 | RA-HMD (Qwen2-VL-7B) | 71.1 | 64.8 | Robust Adaptation of Large Multimodal Models for Retrieval Augmented Hateful Meme Detection | JingbiaoMei/RGCL | 2025 |
| 2 | RGCL | 67.1 | 58.1 | Improving Hateful Meme Detection through Retrieval-Guided Contrastive Learning | JingbiaoMei/RGCL | 2023 |
| 3 | HateCLIPper | 62.4 | 54.8 | Hate-CLIPper: Multimodal Hateful Meme Classification based on Cross-modal Interaction of CLIP Features | gokulkarthik/hateclipper | 2022 |
| 3 | CLIP | 62.4 | 48.1 | Learning Transferable Visual Models From Natural Language Supervision | openai/CLIP · mlfoundations/open_clip · towhee-io/towhee · +79 | 2021 |
| 5 | MultiOFF | – | 50.0 | Multimodal Meme Dataset (MultiOFF) for Identifying Offensive Content in Image and Text | bharathichezhiyan/Multimodal-Meme-Classification-Identifying-Offensive-Content-in-Image-and-Text | 2020 |