gradientsmith
← leaderboard

gpt-oss-20b

200 rollouts

pass@1 clean
69%
pass@1 adversarial
61%
robustness gap
8%
p50 latency
950 ms
p95 latency
2200 ms
cost per attempt
$0.00085
cost per solved clean
$0.0012
cost per solved adv
$0.0014

Per-task breakdown

taskfamilycleanadversarial
code/merge-intervals/001codepasspass
code/max-subarray-sum/001codefailfail
code/roman-to-int/001codepasspass
structured/receipt-cents/001structuredpasspass
structured/apache-log-extract/001structuredpasspass