Multi-modal sarcasm detection on MMSD
93.97AccuracyRoBERTa
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| RoBERTaModality=Text2023.07 | 93.97 | 90.39 | 94.59 | 92.45 | |
| multi-view CLIPModality=Multi-modal, Backbone=clip-vit-base-patch322023.07 | 88.33 | 82.66 | 88.65 | 85.55 | |
| HKEModality=Multi-modal, External Knowledge=false2023.07 | 87.36 | 81.84 | 86.48 | 84.09 | |
| CMGCNModality=Multi-modal, External Knowledge=false2023.07 | 86.54 | — | — | 82.73 | |
| InCrossMGsModality=Multi-modal2023.07 | 86.1 | 81.38 | 84.36 | 82.84 | |
| Att-BERTModality=Multi-modal2023.07 | 86.05 | 80.87 | 85.08 | 82.92 | |
| D&R NetModality=Multi-modal2023.07 | 84.02 | 77.97 | 83.42 | 80.6 | |
| HFMModality=Multi-modal2023.07 | 83.44 | 76.57 | 84.15 | 80.18 | |
| ProCritProtocol=Fine-tuned, Perspective=Adaptive2026.05 | 82.6 | 71.2 | 93.2 | 80.7 | |
| Bi-LSTMModality=Text2023.07 | 81.9 | 76.66 | 78.42 | 77.53 | |
| ProCritProtocol=Fine-tuned, Perspective=Adaptive, Generation Mode=Draft-only2026.05 | 81 | 70.4 | 88.4 | 78.4 | |
| SMSDModality=Text2023.07 | 80.9 | 76.46 | 75.18 | 75.82 | |
| Fixed-role ReasonerProtocol=Fine-tuned, Perspective=Predefined2026.05 | 80.1 | 68.4 | 90.8 | 78 | |
| TextCNNModality=Text2023.07 | 80.03 | 74.29 | 76.39 | 75.32 | |
| Plain CoT ReasonerProtocol=Fine-tuned, Perspective=None2026.05 | 79.4 | 67.5 | 90.8 | 77.5 | |
| S3AgentProtocol=Zero-shot, Perspective=Predefined2026.05 | 73.9 | 69.6 | 58.7 | 63.7 | |
| Zero-shot CoTProtocol=Zero-shot, Perspective=None2026.05 | 73.2 | 68.1 | 59 | 63.2 | |
| Automatic Prompt EngineerProtocol=Zero-shot, Perspective=None2026.05 | 72.9 | 68.9 | 55.8 | 61.6 | |
| Generated Knowledge PromptingProtocol=Zero-shot, Perspective=None2026.05 | 72.6 | 72.2 | 48.1 | 57.8 | |
| Plan-and-SolveProtocol=Zero-shot, Perspective=None2026.05 | 71.9 | 70.6 | 48 | 57.1 | |
| ProCritProtocol=Zero-shot, Perspective=Adaptive2026.05 | 70.9 | 64.5 | 69.1 | 66.7 | |
| ViTModality=Image2023.07 | 67.83 | 57.93 | 70.07 | 63.4 | |
| IRONICProtocol=Zero-shot, Perspective=Predefined2026.05 | 66.8 | 71.4 | 29.6 | 41.8 | |
| Commander-GPTProtocol=Zero-shot, Perspective=Predefined2026.05 | 66 | 55 | 70.1 | 61.6 | |
| ResNetModality=Image2023.07 | 64.76 | 54.41 | 70.8 | 61.53 |