Causal Reasoning on EXTENDED CORR2CAUSE d = 7–24
84.8ParityA-CBO
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| A-CBOTier=HIGH, Base Model=GLM-5.1⋆2026.05 | 84.8 | 84.1 | 81.2 | 80.2 | 75.2 | 86 | 81.9 | 83.2 | |
| A-CBOTier=HIGH, Base Model=Qwen3-30B⋆2026.05 | 83.9 | 83.2 | 80 | 79.9 | 74.2 | 84.9 | 81 | 82.3 | |
| A-CBOTier=MID, Base Model=Qwen3.5-122B2026.05 | 76.6 | 75.9 | 73.1 | 71.5 | 67.1 | 77.8 | 73.7 | 75.1 | |
| A-CBOTier=MID, Base Model=Llama-3.3-70B2026.05 | 73.1 | 72.2 | 69.3 | 68.1 | 63 | 73.8 | 69.9 | 71.3 | |
| A-CBOTier=LOW, Base Model=Gemma-3-12B2026.05 | 59.2 | 58 | 55.4 | 54.2 | 50 | 60.3 | 56.2 | 57.9 | |
| A-CBOTier=LOW, Base Model=LLaMA-7B†2026.05 | 55.4 | 54.6 | 52 | 51.1 | 47.6 | 56.4 | 52.9 | 54.6 |