Pairwise Causal Discovery (test)
90AccuracyFinetuned
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| FinetunedEvaluation Protocol=fine-tuned, Backbone=Mistral-7B-v0.2, Tuning method=LoRA, LoRA rank=8, LoRA alpha=32, Optimizer=AdamW, Epochs=12023.12 | 90 | 58 | 100 | 100 | 100 | |
| GPT-4 turboEvaluation Protocol=zero-shot2023.12 | 82 | 66 | 98 | 87 | 75 | |
| GPT-4Evaluation Protocol=zero-shot2023.12 | 76 | 60 | 100 | 98 | 44 | |
| Mistral-7BEvaluation Protocol=zero-shot, Backbone=Mistral-7B-v0.22023.12 | 75 | 38 | 94 | 92 | 77 | |
| LLaMa2-13BEvaluation Protocol=zero-shot2023.12 | 74 | 8 | 100 | 100 | 86 | |
| Claude 2Evaluation Protocol=zero-shot2023.12 | 74 | 36 | 99 | 84 | 78 | |
| GPT-3.5Evaluation Protocol=zero-shot2023.12 | 58 | 3 | 100 | 96 | 33 |