Counterfactual Reasoning on CRAFT Hard Split (test)
83.64AccuracyCRCG_GPT4
Evaluation Results
| Method | Links | |
|---|---|---|
| CRCG_GPT4Learning Protocol=Few-shot, Proxy Simulator=GPT-42025.06 | 83.64 | |
| GPT-4 with CRCG guided promptLearning Protocol=Few-shot, Prompt Engineering=Causal graph guided2025.06 | 81.22 | |
| GPT-4Learning Protocol=Few-shot, Model Version=gpt-4-06132025.06 | 81.2 | |
| BERT-DLearning Protocol=Supervised (34K instances)2025.06 | 79.34 | |
| CRCG_GPT3.5Learning Protocol=Few-shot, Proxy Simulator=GPT-3.52025.06 | 68.48 | |
| GPT-3.5 with CRCG guided promptLearning Protocol=Few-shot, Prompt Engineering=Causal graph guided2025.06 | 65.89 | |
| LSTM-DLearning Protocol=Supervised (34K instances)2025.06 | 56 | |
| GPT-3.5Learning Protocol=Few-shot, Model Version=gpt-3.5-turbo-06132025.06 | 52.53 |