Code Generation on Wave-4 Wave-8 Qwen-2.5-7B (test)
22.2AchievedDeployed ES
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Deployed ESThreshold=T=peak+32026.06 | 22.2 | — | 51 | 49 | |
| MORBO-proxyHPO=random Pareto HPO2026.06 | 20.1 | 32 | 58 | 40.6 | |
| K-dim CAREGate=A3 + cpp_acc gate2026.06 | 15.1 | 33.5 | 54 | 36.9 | |
| CARE v2Gate Strategy=gate-adjusted2026.06 | 13.8 | 35 | 55 | 47.9 | |
| NaiveBudget=full budget2026.06 | 11.8 | 34.2 | 50 | 37.8 |