Causal Discovery on MECD (test)
2.05Avg SHDDeductive Reasoning
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Deductive ReasoningParadigm=Humans2024.09 | 2.05 | 87.2 | |
| VGCMParadigm=Fine-tuned Ours, causal_inference=true2024.09 | 4.19 | 71.2 | |
| VGCM+Paradigm=Fine-tuned Ours, causal_inference=false2024.09 | 4.51 | 67 | |
| GPT-4oParadigm=Few-shot VLLM Base2024.09 | 4.69 | 65.5 | |
| VideoLLaVAParadigm=Fine-tuned VLLM Base2024.09 | 4.73 | 67.1 | |
| CLIP (ViT-L/14)Paradigm=Fine-tuned Multi-modal2024.09 | 4.77 | 62.9 | |
| VideoChat2Paradigm=Fine-tuned VLLM Base2024.09 | 4.77 | 66.9 | |
| VideoLLaVAParadigm=Few-shot VLLM Base2024.09 | 4.85 | 62.5 | |
| VideoChat2Paradigm=Few-shot VLLM Base2024.09 | 4.89 | 60.7 | |
| Gemini-1.5-ProParadigm=Few-shot LLM Base2024.09 | 4.91 | 59.3 | |
| GPT-4-0613Paradigm=Few-shot LLM Base2024.09 | 4.92 | 59.6 | |
| VideobertParadigm=Fine-tuned Multi-modal2024.09 | 4.95 | 60.9 | |
| VARParadigm=Fine-tuned Multi-modal2024.09 | 4.96 | 57.3 | |
| Video-llamaParadigm=Few-shot VLLM Base2024.09 | 5.1 | 60.6 | |
| MiniGPT-4Paradigm=Few-shot VLLM Base2024.09 | 5.14 | 57.5 | |
| MiniGPT4-videoParadigm=Few-shot VLLM Base2024.09 | 5.16 | 56.8 | |
| Guess all non-causalParadigm=Random Guess2024.09 | 5.36 | 57.6 | |
| Guess all causalParadigm=Random Guess2024.09 | 6.95 | 42.4 |