Multi-entity Reasoning on MEBench Set3 (>100)
94.6Comparison AccuracyDocSage
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| DocSage2026.03 | 94.6 | 88.4 | 79.1 | 87.9 | |
| GPT-4o + RAGRAG=true2026.03 | 50.8 | 35 | 41.3 | 41.5 | |
| StructRAG2026.03 | 49.2 | 37.4 | 35.9 | 40.6 | |
| GraphRAG2026.03 | 45 | 34.4 | 41.7 | 39.6 | |
| GPT-4oRAG=false2026.03 | 15.3 | 21.4 | 30.6 | 21.9 |