| outcomes | correct 206 · incorrect 41 · compile_fail 16 · error 3 · in_progress 1 |
| avg speedup | 4901.13x (over 206 correct) |
| avg memory ratio (kernel / ref) | 0.90x (206 samples) |
| avg fusion ratio (ref launches / kernel launches) | 2331.93x (206 samples) |
| avg energy ratio (ref / kernel) | 16.17x (46 samples) |
| avg SOL score (0–1) | 0.254 (206 samples) |
| avg DRAM util % | 20.3% |
| avg occupancy % | 105.0% |
| avg max abs error | 6.10e-03 |
| model | n | correct | compiled | avg speedup |
|---|---|---|---|---|
| FW-GLM-5-1 | 89 | 68 | 83 | 2892.95x |
| gpt-5.5-priority | 89 | 87 | 89 | 7850.00x |
| grok-4.3 | 89 | 51 | 75 | 2548.29x |