Question Answering on SciQA
76.2Normalized AccuracyGatedFWA-NSA
Evaluation Results
| Method | Links | |
|---|---|---|
| GatedFWA-NSAClass=Attention, Implementation=Triton, Linear=true2025.12 | 76.2 | |
| HGRN2Class=RNN-Like, Implementation=Triton, Linear=true2025.12 | 75.6 | |
| GatedFWAClass=Attention, Implementation=Triton, Linear=true2025.12 | 74.86 | |
| Transformer (LLaMA) + SWA + NSAClass=Attention, Implementation=Triton, Linear=true2025.12 | 74.75 | |
| DeltaNetClass=RNN-Like, Implementation=Triton, Linear=true2025.12 | 74.3 | |
| Transformer (LLaMA)Class=Attention, Implementation=CUDA, Linear=false2025.12 | 73.9 | |
| Transformer (LLaMA) + SWAClass=Attention, Implementation=CUDA, Linear=true2025.12 | 73.79 | |
| MambaClass=RNN-Like, Implementation=CUDA, Linear=true2025.12 | 73.7 | |
| GLAClass=RNN-Like, Implementation=Triton, Linear=true2025.12 | 73.2 | |
| RetNetClass=RNN-Like, Implementation=CUDA, Linear=true2025.12 | 73.1 |