Multimodal Sarcasm Detection on MMSD 2.0 (test)
87.4AccuracyCommander
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| CommanderModel=BERT+ViT, Parameters=110M2025.06 | 87.4 | 86.1 | 85.3 | 85.8 | |
| GDCNet2026.01 | 87.38 | 83.39 | 89.51 | 86.34 | |
| BERT (Fine-tuned)Model=BERT+ViT, Parameters=110M2025.06 | 79.7 | 77.4 | 75.8 | 77.2 | |
| CommanderModel=GPT-4o, Parameters=300B2025.06 | 79.7 | 74.6 | 79.4 | 76.5 | |
| S^3 AgentModel=GPT-4o, Parameters=300B2025.06 | 77.8 | 70.3 | 76.4 | 73.2 | |
| Generated Knowledge PromptingModel=GPT-4o, Parameters=300B2025.06 | 76.4 | 68.9 | 75.1 | 71.8 | |
| Automatic Prompt EngineerModel=GPT-4o, Parameters=300B2025.06 | 75.6 | 67.1 | 73.8 | 70.3 | |
| GPT-4oPrompting strategy=CoT2026.01 | 74.26 | 65.81 | 72.68 | 68.92 | |
| Zero-shot CoTModel=GPT-4o, Parameters=300B2025.06 | 74.2 | 65.8 | 72.6 | 68.9 | |
| CommanderModel=MiniCPM-V2, Parameters=2.8B2025.06 | 73.9 | 66.8 | 78.1 | 72 | |
| CommanderModel=Gemini Pro, Parameters=600B2025.06 | 73.8 | 68.4 | 76.7 | 71.9 | |
| Plan-and-SolveModel=GPT-4o, Parameters=300B2025.06 | 73.1 | 64.2 | 71 | 67.4 | |
| GPT-4oPrompting strategy=Zero-Shot2026.01 | 71.07 | 79.52 | 71.07 | 70.24 | |
| CommanderModel=Yi-VL, Parameters=6B2025.06 | 67.9 | 69 | 46.1 | 55.3 | |
| CommanderModel=Qwen-VL-Chat, Parameters=9B2025.06 | 67.7 | 58.8 | 83.2 | 68.9 | |
| S^3 AgentModel=Gemini Pro, Parameters=600B2025.06 | 66.5 | 58 | 80.3 | 67.4 | |
| S^3 AgentModel=DeepSeek-VL-Chat, Parameters=7B2025.06 | 64.9 | 63.3 | 44.1 | 52 | |
| S^3 AgentModel=Qwen-VL-Chat, Parameters=9B2025.06 | 63.3 | 57 | 60.4 | 58.6 | |
| Plan-and-SolveModel=DeepSeek-VL-Chat, Parameters=7B2025.06 | 63 | 60.5 | 40.2 | 48.3 | |
| S^3 AgentModel=MiniCPM-V2, Parameters=2.8B2025.06 | 62.9 | 60 | 41.8 | 49.2 | |
| Zero-shot CoTModel=DeepSeek-VL-Chat, Parameters=7B2025.06 | 62.8 | 59.8 | 41.3 | 48.8 | |
| Automatic Prompt EngineerModel=DeepSeek-VL-Chat, Parameters=7B2025.06 | 62.3 | 59.7 | 38.2 | 46.6 | |
| Generated Knowledge PromptingModel=Gemini Pro, Parameters=600B2025.06 | 61.7 | 53.6 | 82.8 | 65.1 | |
| Zero-shot CoTModel=Gemini Pro, Parameters=600B2025.06 | 61.1 | 53.9 | 66.1 | 59.4 | |
| Automatic Prompt EngineerModel=Gemini Pro, Parameters=600B2025.06 | 60.4 | 53.1 | 68.1 | 60 | |
| Qwen-VLPrompting strategy=CoT2026.01 | 58.86 | 56.82 | 58.67 | 57.26 | |
| Generated Knowledge PromptingModel=DeepSeek-VL-Chat, Parameters=7B2025.06 | 58.1 | 53.8 | 18.5 | 27.5 | |
| Plan-and-SolveModel=Gemini Pro, Parameters=600B2025.06 | 56.5 | 49.6 | 58.5 | 53.7 | |
| S^3 AgentModel=Yi-VL, Parameters=6B2025.06 | 54.3 | 44.4 | 24.6 | 31.7 | |
| Zero-shot CoTModel=MiniCPM-V2, Parameters=2.8B2025.06 | 51.6 | 40.9 | 27.7 | 33 | |
| LLaVAPrompting strategy=Zero-Shot2026.01 | 51.06 | 40.09 | 46.4 | 43.02 | |
| LLaVAPrompting strategy=CoT2026.01 | 48.69 | 40.93 | 65.17 | 50.28 | |
| Automatic Prompt EngineerModel=MiniCPM-V2, Parameters=2.8B2025.06 | 48 | 36.2 | 27.4 | 31.2 | |
| Generated Knowledge PromptingModel=MiniCPM-V2, Parameters=2.8B2025.06 | 47.3 | 41.6 | 55.3 | 47.5 | |
| Plan-and-SolveModel=MiniCPM-V2, Parameters=2.8B2025.06 | 46.8 | 38 | 37.4 | 37.7 | |
| CommanderModel=DeepSeek-VL-Chat, Parameters=7B2025.06 | 46.7 | 44.4 | 94.7 | 60.5 | |
| Qwen-VLPrompting strategy=Zero-Shot2026.01 | 40.63 | 32.44 | 35.53 | 33.63 | |
| Automatic Prompt EngineerModel=Qwen-VL-Chat, Parameters=9B2025.06 | 40.4 | 32.1 | 34.5 | 33.3 | |
| Zero-shot CoTModel=Qwen-VL-Chat, Parameters=9B2025.06 | 40 | 32 | 35.3 | 33.6 | |
| Generated Knowledge PromptingModel=Qwen-VL-Chat, Parameters=9B2025.06 | 38.9 | 38.4 | 37.6 | 28 | |
| Plan-and-SolveModel=Qwen-VL-Chat, Parameters=9B2025.06 | 38.7 | 32.1 | 38 | 34.8 | |
| Zero-shot CoTModel=Yi-VL, Parameters=6B2025.06 | 16.2 | 6 | 6.3 | 6.1 | |
| Automatic Prompt EngineerModel=Yi-VL, Parameters=6B2025.06 | 14.2 | 4.2 | 4.5 | 4.3 | |
| Generated Knowledge PromptingModel=Yi-VL, Parameters=6B2025.06 | 11.2 | 4 | 4.6 | 4.3 | |
| Plan-and-SolveModel=Yi-VL, Parameters=6B2025.06 | 9.2 | 4 | 4.8 | 4.4 |