Commonsense Reasoning on HellaSwag, OBQA, WinoGrande, ARC-c, ARC-e, PIQA
52.3HellaSwagYOCO++
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| YOCO++Evaluation Protocol=Zero-shot2026.04 | 52.3 | 34 | 55.48 | 27.9 | 53.11 | 71.16 | 48.99 | |
| FusedKV-LiteEvaluation Protocol=Zero-shot2026.04 | 52.28 | 32.8 | 54.93 | 28.24 | 52.19 | 70.13 | 48.43 | |
| FusedKVEvaluation Protocol=Zero-shot2026.04 | 52.04 | 32 | 55.48 | 28.67 | 51.98 | 70.35 | 48.42 | |
| TransformerEvaluation Protocol=Zero-shot2026.04 | 51.25 | 33 | 56.04 | 27.56 | 52.1 | 70.29 | 48.37 | |
| YOCOEvaluation Protocol=Zero-shot2026.04 | 51.21 | 32 | 53.59 | 29.01 | 51.43 | 70.62 | 47.98 |