Reasoning on BBEH (test)
34.5AccuracyCorrectness Only
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Correctness OnlyModel=Qwen3-14b2026.02 | 34.5 | 53.8 | |
| REMULModel=Qwen3-14b2026.02 | 33.3 | 59.3 | |
| MAT-SteerModel=Qwen3-14b2026.02 | 32.6 | 58.1 | |
| OriginalModel=Qwen3-14b2026.02 | 31.3 | 56.7 | |
| Correctness OnlyModel=Phi4-reasoning-plus2026.02 | 29.1 | 44.1 | |
| Balanced RewardsModel=Qwen3-14b2026.02 | 28.9 | 52.4 | |
| Faithfulness OnlyModel=Qwen3-14b2026.02 | 28.4 | 61.8 | |
| REMULModel=Phi4-reasoning-plus2026.02 | 28.2 | 48.6 | |
| MAT-SteerModel=Phi4-reasoning-plus2026.02 | 26.9 | 46 | |
| OriginalModel=Phi4-reasoning-plus2026.02 | 26.6 | 45.2 | |
| Hint OptimizedModel=Qwen3-14b2026.02 | 26.1 | 71.2 | |
| Balanced RewardsModel=Phi4-reasoning-plus2026.02 | 25.8 | 44.7 | |
| Faithfulness OnlyModel=Phi4-reasoning-plus2026.02 | 23.5 | 49.8 | |
| Hint OptimizedModel=Phi4-reasoning-plus2026.02 | 21.8 | 53.4 |