pop_l123_popcorn2_bench_all_gpt · variant overview
trajectories
102
finished
101
correct
20
compiled
52
avg speedup (correct)
3341.83x
sweep-wide eval metricsaveraged over correct kernels
outcomeserror 49 · incorrect 32 · correct 20 · in_progress 1
avg speedup3341.83x (over 20 correct)
avg memory ratio (kernel / ref)0.94x (20 samples)
avg fusion ratio (ref launches / kernel launches)6547.08x (20 samples)
avg energy ratio (ref / kernel)27.18x (3 samples)
avg SOL score (0–1)0.214 (20 samples)
avg DRAM util %15.5%
avg occupancy %145.7%
avg max abs error2.24e-06
modelncorrectcompiled avg speedup
FW-GLM-5-1 34 10 24 1865.25x
gpt-5.5-priority 34 10 28 4405.41x
grok-4.3 34 0 0 0.00x