Commonsense Reasoning on HellaSWAG (Normalized Accuracy)
63.84HellaSWAG Normalized AccuracyGated KalmaNet
Evaluation results
| Method | Links | |
|---|---|---|
| Gated KalmaNetevaluation_mode=zero-shot, parameters=2.8B, context_length=< 2K tokens2025.11 | 63.84 | |
| Gated DeltaNetevaluation_mode=zero-shot, parameters=2.8B, context_length=< 2K tokens2025.11 | 62.8 | |
| Mamba2evaluation_mode=zero-shot, parameters=2.8B, context_length=< 2K tokens2025.11 | 62.23 | |
| DeltaNetevaluation_mode=zero-shot, parameters=2.8B, context_length=< 2K tokens2025.11 | 61.13 | |
| Transformerevaluation_mode=zero-shot, parameters=2.8B, context_length=< 2K tokens2025.11 | 60.96 | |
| Gated Linear Attentionevaluation_mode=zero-shot, parameters=2.8B, context_length=< 2K tokens2025.11 | 48.83 |