Question Answering on MuSiQue Covered Questions (Acc, F1, EM)
53.2AccuracySPARQL
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| SPARQLModel=70B, GW=false2026.03 | 53.2 | 40.4 | 29.6 | |
| SPARQL + GWModel=70B, GW=true2026.03 | 51.4 | 42.3 | 32.5 | |
| Base + GWModel=70B, GW=true2026.03 | 46.2 | 37.3 | 26.2 | |
| BaseModel=70B, GW=false2026.03 | 40.3 | 32.5 | 23.9 | |
| SPARQL + GWModel=8B, GW=true2026.03 | 35.1 | 26.4 | 20 | |
| SPARQLModel=8B, GW=false2026.03 | 34.8 | 27.4 | 19.7 | |
| BaseModel=8B, GW=false2026.03 | 27.5 | 21.2 | 15.8 | |
| Base + GWModel=8B, GW=true2026.03 | 22.6 | 17 | 11.4 |