Causal Discovery on Real-world corpora
87.8F1GSCITE
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| SCITEn=32026.04 | 87.8 | 0.48 | 0.15 | |
| SCITEn=42026.04 | 82.3 | 1.1 | 0.45 | |
| LLM-CGn=32026.04 | 81.5 | 0.78 | 0.27 | |
| SCITEn=52026.04 | 80 | 1.69 | 0.86 | |
| LLM-CGn=42026.04 | 77.4 | 1.66 | 0.77 | |
| SCITEn=62026.04 | 73.8 | 2.38 | 1.45 | |
| RuleBayesn=32026.04 | 71.6 | 1.35 | 0.51 | |
| SCITEn=72026.04 | 71.6 | 3.11 | 2.1 | |
| LLM-CGn=52026.04 | 70.1 | 2.64 | 1.51 | |
| LLM-CGn=62026.04 | 67.6 | 3.61 | 2.33 | |
| SCITEn=82026.04 | 67.4 | 3.91 | 2.89 | |
| RuleBayesn=42026.04 | 65.9 | 2.77 | 1.51 | |
| LLM-CGn=72026.04 | 62.8 | 4.45 | 3.36 | |
| SCITEn=92026.04 | 61.9 | 4.65 | 3.72 | |
| RuleBayesn=52026.04 | 61.4 | 4.58 | 2.88 | |
| SCITEn=102026.04 | 58.5 | 5.46 | 4.57 | |
| RuleBayesn=62026.04 | 55.3 | 6.11 | 4.79 | |
| LLM-CGn=82026.04 | 55.3 | 5.48 | 4.39 | |
| LLM-CGn=92026.04 | 50.7 | 6.42 | 5.75 | |
| RuleBayesn=72026.04 | 50.3 | 7.94 | 6.88 | |
| LLM-CGn=102026.04 | 47.2 | 7.79 | 7.08 | |
| RuleBayesn=82026.04 | 44.6 | 9.39 | 9.96 | |
| RuleBayesn=92026.04 | 39 | 11.56 | 12.9 | |
| RuleBayesn=102026.04 | 35.7 | 13.2 | 16.7 | |
| SCITEn=3, Backbone=DeepSeek-R12026.04 | 0.878 | 0.48 | 0.15 | |
| SCITEn=32026.04 | 0.878 | 0.48 | 0.15 | |
| SCITEn=4, Backbone=DeepSeek-R12026.04 | 0.823 | 1.1 | 0.45 | |
| SCITEn=42026.04 | 0.823 | 1.1 | 0.45 | |
| LLM-CGn=3, Backbone=DeepSeek-R12026.04 | 0.815 | 0.78 | 0.27 | |
| LLM-CGn=3, model=Qwen3-235B-A22B-Thinking-25072026.04 | 0.815 | 0.78 | 0.27 | |
| SCITEn=5, Backbone=DeepSeek-R12026.04 | 0.8 | 1.69 | 0.86 | |
| SCITEn=52026.04 | 0.8 | 1.69 | 0.86 | |
| LLM-CGn=4, Backbone=DeepSeek-R12026.04 | 0.774 | 1.66 | 0.77 | |
| LLM-CGn=4, model=Qwen3-235B-A22B-Thinking-25072026.04 | 0.774 | 1.66 | 0.77 | |
| SCITEn=6, Backbone=DeepSeek-R12026.04 | 0.738 | 2.38 | 1.45 | |
| SCITEn=62026.04 | 0.738 | 2.38 | 1.45 | |
| RuleBayesn=3, Backbone=DeepSeek-R12026.04 | 0.716 | 1.35 | 0.51 | |
| SCITEn=7, Backbone=DeepSeek-R12026.04 | 0.716 | 3.11 | 2.1 | |
| RuleBayesn=32026.04 | 0.716 | 1.35 | 0.51 | |
| SCITEn=72026.04 | 0.716 | 3.11 | 2.1 | |
| LLM-CGn=5, Backbone=DeepSeek-R12026.04 | 0.701 | 2.64 | 1.51 | |
| LLM-CGn=5, model=Qwen3-235B-A22B-Thinking-25072026.04 | 0.701 | 2.64 | 1.51 | |
| LLM-CGn=6, Backbone=DeepSeek-R12026.04 | 0.676 | 3.61 | 2.33 | |
| LLM-CGn=6, model=Qwen3-235B-A22B-Thinking-25072026.04 | 0.676 | 3.61 | 2.33 | |
| SCITEn=8, Backbone=DeepSeek-R12026.04 | 0.674 | 3.91 | 2.89 | |
| SCITEn=82026.04 | 0.674 | 3.91 | 2.89 | |
| RuleBayesn=4, Backbone=DeepSeek-R12026.04 | 0.659 | 2.77 | 1.51 | |
| RuleBayesn=42026.04 | 0.659 | 2.77 | 1.51 | |
| LLM-CGn=7, Backbone=DeepSeek-R12026.04 | 0.628 | 4.45 | 3.36 | |
| LLM-CGn=7, model=Qwen3-235B-A22B-Thinking-25072026.04 | 0.628 | 4.45 | 3.36 | |
| SCITEn=9, Backbone=DeepSeek-R12026.04 | 0.619 | 4.65 | 3.72 | |
| SCITEn=92026.04 | 0.619 | 4.65 | 3.72 | |
| RuleBayesn=5, Backbone=DeepSeek-R12026.04 | 0.614 | 4.58 | 2.88 | |
| RuleBayesn=52026.04 | 0.614 | 4.58 | 2.88 | |
| SCITEn=10, Backbone=DeepSeek-R12026.04 | 0.585 | 5.46 | 4.57 | |
| SCITEn=102026.04 | 0.585 | 5.46 | 4.57 | |
| RuleBayesn=6, Backbone=DeepSeek-R12026.04 | 0.553 | 6.11 | 4.79 | |
| LLM-CGn=8, Backbone=DeepSeek-R12026.04 | 0.553 | 5.48 | 4.39 | |
| RuleBayesn=62026.04 | 0.553 | 6.11 | 4.79 | |
| LLM-CGn=8, model=Qwen3-235B-A22B-Thinking-25072026.04 | 0.553 | 5.48 | 4.39 | |
| LLM-CGn=9, Backbone=DeepSeek-R12026.04 | 0.507 | 6.42 | 5.75 | |
| LLM-CGn=9, model=Qwen3-235B-A22B-Thinking-25072026.04 | 0.507 | 6.42 | 5.75 | |
| RuleBayesn=7, Backbone=DeepSeek-R12026.04 | 0.503 | 7.94 | 6.88 | |
| RuleBayesn=72026.04 | 0.503 | 7.94 | 6.88 | |
| LLM-CGn=10, Backbone=DeepSeek-R12026.04 | 0.472 | 7.79 | 7.08 | |
| LLM-CGn=10, model=Qwen3-235B-A22B-Thinking-25072026.04 | 0.472 | 7.79 | 7.08 | |
| RuleBayesn=8, Backbone=DeepSeek-R12026.04 | 0.446 | 9.39 | 9.96 | |
| RuleBayesn=82026.04 | 0.446 | 9.39 | 9.96 | |
| RuleBayesn=9, Backbone=DeepSeek-R12026.04 | 0.39 | 11.56 | 12.9 | |
| RuleBayesn=92026.04 | 0.39 | 11.56 | 12.9 | |
| RuleBayesn=10, Backbone=DeepSeek-R12026.04 | 0.357 | 13.2 | 16.7 | |
| RuleBayesn=102026.04 | 0.357 | 13.2 | 16.7 |