Multi-entity Reasoning on MEBench All sets
93.4Comparison AccDocSage
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| DocSage2026.03 | 93.4 | 90.8 | 81.2 | 89.2 | |
| GPT-4o + RAGRAG=true2026.03 | 69.6 | 57.9 | 59.3 | 62 | |
| StructRAG2026.03 | 67.8 | 58.8 | 57.3 | 61.2 | |
| GraphRAG2026.03 | 61.8 | 55.8 | 59.3 | 58.6 | |
| GPT-4oRAG=false2026.03 | 26.2 | 35.3 | 40.7 | 33.8 |