Causal Discovery on CausalMan Medium (test)
27.4F1 ScoreCAST
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| CASTLLM=Qwen 3 14B, Variant=HITL+RAG2026.07 | 27.4 | 65.6 | 17.5 | 558 | 4,179.19 | |
| CASTLLM=Qwen 3 14B, Variant=HITL2026.07 | 26.6 | 50.8 | 19.1 | 610.8 | 5,905.02 | |
| CASTLLM=Qwen 3 14B, Variant=RAG2026.07 | 21.1 | 48.5 | 13.9 | 568.4 | 4,901.53 | |
| CASTLLM=o3-mini, Variant=RAG2026.07 | 20.9 | 51.8 | 13.1 | 571.6 | 1,321.15 | |
| CASTLLM=o3-mini, Variant=HITL2026.07 | 17.5 | 49.9 | 10.8 | 558.8 | 1,052.11 | |
| CASTLLM=o3-mini, Variant=HITL+RAG2026.07 | 14.1 | 35.7 | 8.9 | 579.6 | 1,370.25 | |
| CASTLLM=gpt 4o-mini, Variant=RAG2026.07 | 11.3 | 29.6 | 7.2 | 594.2 | 1,241.17 | |
| CASTLLM=gpt 4o-mini, Variant=HITL2026.07 | 10.3 | 39.2 | 5.6 | 547 | 1,929.17 | |
| CASTLLM=gpt 4o-mini, Variant=HITL+RAG2026.07 | 8.6 | 27.1 | 5.1 | 568.2 | 2,488.79 | |
| BOSS2026.07 | 5.7 | 3 | — | 1,817.4 | — | |
| XGES2026.07 | 1.4 | 21.1 | 0.7 | 547 | — | |
| CausalCopilot2026.07 | 1.4 | 21.1 | 0.7 | 547 | — | |
| GranDAG2026.07 | 0 | 0 | 0 | 544.4 | 1,061.17 |