Image-based Question Answering on H-Meme (Hateful Meme)
69.1AUCAnyMAL 70B (ViT-G)
Evaluation Results
| Method | Links | |
|---|---|---|
| AnyMAL 70B (ViT-G)LLM Size=70B, Vision Encoder=ViT-G, Zeroshot=true2023.09 | 69.1 | |
| AnyMAL 70B (ViT-L)LLM Size=70B, Vision Encoder=ViT-L, Zeroshot=true2023.09 | 68.2 | |
| AnyMAL 70B (MM-IT; ViT-G)LLM Size=70B, Vision Encoder=ViT-G, Instruction-Tuned=true2023.09 | 67.4 | |
| AnyMAL 13B (ViT-G)LLM Size=13B, Vision Encoder=ViT-G, Zeroshot=true2023.09 | 66 | |
| AnyMAL 70B (DINO-v2)LLM Size=70B, Vision Encoder=DINO-v2, Zeroshot=true2023.09 | 65.6 | |
| IBELICS-80BLLM Size=80B, Zeroshot=true2023.09 | 60.6 | |
| Flamingo-9BLLM Size=9B, Zeroshot=true2023.09 | 57 | |
| Flamingo-3BLLM Size=3B, Zeroshot=true2023.09 | 53.7 | |
| InstructBLIP (V-13B)LLM Size=13B, Zeroshot=true2023.09 | 53.7 | |
| BLIP-2 (FlanT5_XXL)Zeroshot=true2023.09 | 52 | |
| IBELICS-9BLLM Size=9B, Zeroshot=true2023.09 | 51.8 | |
| OpenFlamingo-v2Zeroshot=true2023.09 | 51.6 | |
| Flamingo-80BLLM Size=80B, Zeroshot=true2023.09 | 46.4 |