Mathematical Reasoning on MATH GAOKAO Benchmark 13
0.445NLLRCC
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| RCCModel=Qwen3-8B2026.01 | 0.445 | 3.63 | |
| RCCModel=Gemma3-12B2026.01 | 0.484 | 5.07 | |
| UQAC (Attention Chain)Model=Gemma3-12B2026.01 | 0.503 | 13.35 | |
| Logits (Average)Model=Qwen3-8B2026.01 | 0.558 | 14.04 | |
| Logits (Average)Model=Gemma3-12B2026.01 | 0.56 | 14.25 | |
| SAR (Shifting Attention to Relevance)Model=Gemma3-12B2026.01 | 0.594 | 9.46 | |
| UQAC (Attention Chain)Model=Qwen3-8B2026.01 | 0.729 | 10.07 | |
| SAR (Shifting Attention to Relevance)Model=Qwen3-8B2026.01 | 0.932 | 13.87 | |
| Self-ConsistencyModel=Gemma3-12B2026.01 | 1.163 | 13.71 | |
| Self-ConsistencyModel=Qwen3-8B2026.01 | 1.573 | 12.19 | |
| Logits (Final)Model=Gemma3-12B2026.01 | 2.122 | 18.84 | |
| Self-EvaluationModel=Gemma3-12B2026.01 | 2.704 | 16.57 | |
| Self-EvaluationModel=Qwen3-8B2026.01 | 3.533 | 16.93 | |
| Logits (Final)Model=Qwen3-8B2026.01 | 4.369 | 18.95 |