Question Answering and Reasoning on Downstream Average 0-shot
60.5Average AccuracyAIRAhybrid-D
Evaluation results
| Method | Links | ||
|---|---|---|---|
| AIRAhybrid-DShot=0-shot, Architecture Variant=Stretched2026.05 | 60.5 | 62.2 | |
| Nemotron-2 Approx.Shot=0-shot2026.05 | 60.3 | 61.5 | |
| AIRAhybrid-AShot=0-shot, Architecture Variant=Stretched2026.05 | 60.3 | 62.1 | |
| AIRAhybrid-BShot=0-shot, Architecture Variant=Stretched2026.05 | 60.2 | 61.9 | |
| AIRAhybrid-DShot=0-shot, Architecture Variant=Stacked2026.05 | 60.2 | 61.6 | |
| AIRAhybrid-BShot=0-shot, Architecture Variant=Stacked2026.05 | 60.1 | 61.8 | |
| Composer (2Mb-M-3A)Shot=0-shot2026.05 | 59.9 | 61.6 | |
| AIRAhybrid-CShot=0-shot, Architecture Variant=Stretched2026.05 | 59.9 | 61.2 | |
| AIRAhybrid-EShot=0-shot, Architecture Variant=Stretched2026.05 | 59.8 | 61.7 | |
| Mamba (Mb + M)Shot=0-shot2026.05 | 59.6 | 60.9 | |
| Nemotron-H Approx.Shot=0-shot2026.05 | 59.2 | 60.1 | |
| AIRAhybrid-EShot=0-shot, Architecture Variant=Stacked2026.05 | 59.2 | 61 |