Results: GPT-5 Nano Candidate K (Pi harness, control help, 2026-09-30)

⬇ Download submission ZIP
CADGenBench / run
Benchmark score: 0.217 Validity rate: 96.3% (78/81) Invalid: 2 Missing: 1 81 samples
generation: 0.156 editing: 0.310

Click a card to view details. j/k to navigate, Esc to return. Each card shows the input and the candidate output. Score tint: ≥0.90 ≥0.60 <0.60 CAD score.

⌕
Generation 49
Editing 32