Language Model Evaluation on DCLM Core
49.3DCLM Core ScoreComposer (2Mb-M-3A)
Evaluation Results
| Method | Links | |
|---|---|---|
| Composer (2Mb-M-3A)Shot=Few-shots2026.05 | 49.3 | |
| AIRAhybrid-BShot=Few-shots, Architecture Variant=Stretched2026.05 | 49.1 | |
| AIRAhybrid-BShot=Few-shots, Architecture Variant=Stacked2026.05 | 49 | |
| AIRAhybrid-CShot=Few-shots, Architecture Variant=Stretched2026.05 | 48.9 | |
| AIRAhybrid-EShot=Few-shots, Architecture Variant=Stacked2026.05 | 48.8 | |
| AIRAhybrid-EShot=Few-shots, Architecture Variant=Stretched2026.05 | 48.6 | |
| Nemotron-H Approx.Shot=Few-shots2026.05 | 48.5 | |
| AIRAhybrid-DShot=Few-shots, Architecture Variant=Stretched2026.05 | 48.5 | |
| Nemotron-2 Approx.Shot=Few-shots2026.05 | 48.4 | |
| AIRAhybrid-DShot=Few-shots, Architecture Variant=Stacked2026.05 | 48.4 | |
| AIRAhybrid-AShot=Few-shots, Architecture Variant=Stretched2026.05 | 47.5 | |
| Mamba (Mb + M)Shot=Few-shots2026.05 | 46.1 |