Multi-entity Reasoning on MEBench Set1 (0-10)
96.8Comparison AccuracyDocSage
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| DocSage2026.03 | 96.8 | 92.9 | 83.7 | 91.8 | |
| GPT-4o + RAGRAG=true2026.03 | 87 | 69 | 75.5 | 76.4 | |
| StructRAG2026.03 | 83.8 | 77.3 | 73.5 | 78.4 | |
| GraphRAG2026.03 | 77.4 | 76.1 | 69.4 | 74.8 | |
| GPT-4oRAG=false2026.03 | 46.7 | 59.5 | 57.1 | 54.8 |