Long-range dependency modeling on Long Range Arena 100 tokens
85.5ListOps AccuracyStable-Mamba
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Stable-MambaModel Size=2.8B2026.02 | 85.5 | 70 | 55.5 | 100 | 77.75 | |
| Stable-MambaModel Size=1.4B2026.02 | 83 | 70 | 42.6 | 77 | 68.15 | |
| Steered MambaModel Size=2.8B2026.02 | 81 | 68.5 | 50.5 | 70 | 67.5 | |
| Stable-MambaModel Size=130M2026.02 | 80 | 60 | 37 | 95 | 68 | |
| Steered MambaModel Size=1.4B2026.02 | 78 | 68.1 | 38.2 | 70 | 63.58 | |
| MambaModel Size=2.8B2026.02 | 74 | 69 | 49.5 | 70 | 65.63 | |
| Hyena2026.02 | 74 | 20 | 26 | 100 | 55 | |
| Steered MambaModel Size=130M2026.02 | 73 | 61 | 35.5 | 74 | 60.88 | |
| Mamba-22026.02 | 67 | 38 | 38 | 1 | 36 | |
| MambaModel Size=1.4B2026.02 | 65 | 68 | 26.5 | 65 | 56.13 | |
| GPT-22026.02 | 64 | 20 | 66.2 | 100 | 62.55 | |
| MambaModel Size=130M2026.02 | 62 | 61 | 25 | 60 | 52 | |
| DenseMamba2026.02 | 50 | 40.6 | 43 | 95 | 57.15 | |
| MiniPLM-MambaModel Size=130M2026.02 | 48 | 5 | 74.5 | 4 | 32.88 |