Mathematical Reasoning on MMLU High School Math
94.81Accuracy (Raw consensus)Raw consensus
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| Raw consensusModel=Qwen3-4B, Policy=no-framing majority baseline, Sample count (N)=2702026.05 | 94.81 | — | — | — | — | — | — | |
| ARBITER-∆Model=Qwen3-4B, Policy=fixed framed+guided parameter-free, Sample count (N)=2702026.05 | 94.81 | 95.19 | 0.37 | 10 | 3 | 2 | 1 | |
| Raw consensusModel=Phi-4, Policy=no-framing majority baseline, Sample count (N)=2702026.05 | 92.59 | — | — | — | — | — | — | |
| ARBITER-∆Model=Phi-4, Policy=fixed framed+guided parameter-free, Sample count (N)=2702026.05 | 92.59 | 93.7 | 1.11 | 4 | 3 | 0 | 3 | |
| Raw consensusModel=Llama-3.1-8B, Policy=no-framing majority baseline, Sample count (N)=2702026.05 | 78.52 | — | — | — | — | — | — | |
| ARBITER-∆Model=Llama-3.1-8B, Policy=fixed framed+guided parameter-free, Sample count (N)=2702026.05 | 78.52 | 80 | 1.48 | 17 | 9 | 5 | 4 |