Budget Allocation on Mostly-Hard
27.8AccuracyOracle
Evaluation Results
| Method | Links | |
|---|---|---|
| OracleBudget=4096, Model=Qwen3-30B-A3B-Instruct2026.06 | 27.8 | |
| OracleBudget=1024, Model=Qwen3-30B-A3B-Instruct2026.06 | 17.4 | |
| PredictorBudget=4096, Model=Qwen3-30B-A3B-Instruct2026.06 | 15.2 | |
| CLEAR (Heur.)Budget=4096, Model=Qwen3-30B-A3B-Instruct2026.06 | 15.2 | |
| CLEAR (Ours)Budget=4096, Model=Qwen3-30B-A3B-Instruct2026.06 | 14.8 | |
| UniformBudget=4096, Model=Qwen3-30B-A3B-Instruct2026.06 | 14.2 | |
| CLEAR (Auction)Budget=4096, Model=Qwen3-30B-A3B-Instruct2026.06 | 13 | |
| CLEAR (Heur.)Budget=1024, Model=Qwen3-30B-A3B-Instruct2026.06 | 9.8 | |
| CLEAR (Ours)Budget=1024, Model=Qwen3-30B-A3B-Instruct2026.06 | 8.6 | |
| CLEAR (Auction)Budget=1024, Model=Qwen3-30B-A3B-Instruct2026.06 | 7 | |
| UniformBudget=1024, Model=Qwen3-30B-A3B-Instruct2026.06 | 6.4 | |
| PredictorBudget=1024, Model=Qwen3-30B-A3B-Instruct2026.06 | 1.8 |