Graph Reasoning on GRBENCH Legal
56.7QwenScoreGraphScout
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| GraphScoutLLM=4B w/ train2026.03 | 56.7 | 59.7 | |
| GraphCoTLLM=DeepSeek-Chat2026.03 | 55.5 | 55.9 | |
| GraphScoutLLM=8B w/ train2026.03 | 55.5 | 57.3 | |
| GraphCoTLLM=Qwen-Max2026.03 | 53.3 | 53.1 | |
| GraphCoTLLM=GLM-4.62026.03 | 51.1 | 51.7 | |
| GraphCoTLLM=GPT-4o2026.03 | 49.4 | 40.3 | |
| GraphCounselorLLM=DeepSeek-Chat2026.03 | 48.3 | 28.6 | |
| GraphCounselorLLM=Qwen-Max2026.03 | 47.8 | 26.2 | |
| GraphCounselorLLM=GPT-4o2026.03 | 43.3 | 21.9 | |
| CypherLLM=GLM-4.62026.03 | 42.2 | 10.6 | |
| CypherLLM=Qwen-Max2026.03 | 40.6 | 13.2 | |
| CypherLLM=GPT-4o2026.03 | 38.9 | 11.8 | |
| GraphRAGLLM=DeepSeek-Chat2026.03 | 36.1 | 38.1 | |
| GraphScoutLLM=8B w/o train2026.03 | 36.1 | 36.4 | |
| GraphRAGLLM=Qwen-Max2026.03 | 35.6 | 42.4 | |
| PolyGLLM=GLM-4.62026.03 | 35.6 | 7.9 | |
| CypherLLM=DeepSeek-Chat2026.03 | 35 | 9.4 | |
| TextRAGLLM=Qwen-Max2026.03 | 34.4 | 40.6 | |
| PolyGLLM=GPT-4o2026.03 | 34.4 | 8.7 | |
| PolyGLLM=DeepSeek-Chat2026.03 | 33.9 | 7 | |
| GraphCounselorLLM=GLM-4.62026.03 | 33.9 | 24.6 | |
| TextRAGLLM=DeepSeek-Chat2026.03 | 33.3 | 36.5 | |
| PolyGLLM=Qwen-Max2026.03 | 31.7 | 7.4 | |
| TextRAGLLM=GPT-4o2026.03 | 25.6 | 23.2 | |
| BaseLLMLLM=GPT-4o2026.03 | 24.4 | 11 | |
| BaseLLMLLM=Qwen-Max2026.03 | 24.4 | 29.8 | |
| GraphRAGLLM=GPT-4o2026.03 | 23.9 | 23.3 | |
| BaseLLMLLM=GLM-4.62026.03 | 22.2 | 24.3 | |
| BaseLLMLLM=DeepSeek-Chat2026.03 | 22.2 | 26.5 | |
| TextRAGLLM=GLM-4.62026.03 | 21.7 | 25.3 | |
| GraphScoutLLM=4B w/o train2026.03 | 18.9 | 22.6 | |
| GraphRAGLLM=GLM-4.62026.03 | 17.2 | 25.3 |