Question Answering on StrategyQA (dev)
78.17AccuracyGEEK
Evaluation Results
| Method | Links | |
|---|---|---|
| GEEKModel size=11B2026.03 | 78.17 | |
| KARDModel size=3B2026.03 | 70.55 | |
| GR+RATDModel size=440M2026.03 | 64.2 | |
| StableVicuna INT8Model size=13B2026.03 | 61.7 | |
| UL2Model size=20B2026.03 | 59 |