Model Routing on Dataset A
83.25Accuracy (Selected Answer)3-model oracle
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| 3-model oracle2026.06 | 83.25 | 100 | — | — | — | — | — | |
| Brick maxr=+1.02026.06 | 76.98 | 17.77 | 0.0221 | — | 0 | 31 | 69 | |
| Brick highr=+0.52026.06 | 76.24 | 35.63 | 0.0149 | — | 21 | 35 | 44 | |
| always kimi2026.06 | 75.02 | 21.28 | 0.0307 | — | 0 | 0 | 100 | |
| RouteLLM binary2026.06 | 75.02 | 21.31 | 0.0307 | — | 0 | 0 | 100 | |
| RouteLLM tournament2026.06 | 75.02 | 21.31 | 0.0307 | — | 0 | 0 | 100 | |
| Brick neutralr=02026.06 | 74.11 | 40.35 | 0.0065 | — | 29 | 56 | 15 | |
| always ds42026.06 | 73.69 | 15.55 | 0.0029 | — | 0 | 100 | 0 | |
| Cascade Routingmode=parallel scorer2026.06 | 73.4 | 28.96 | 0.0061 | 0.0061 | 18 | 61 | 21 | |
| Brick lowr=−0.52026.06 | 71.62 | 48.66 | 0.0036 | — | 45 | 51 | 5 | |
| FrugalGPT cascade2026.06 | 69.42 | 31.03 | 0.0028 | 0.0041 | 33 | 64 | 3 | |
| always qwen2026.06 | 63.17 | 63.17 | 0.0014 | — | 100 | 0 | 0 | |
| Brick minr=−1.02026.06 | 63.17 | 63.17 | 0.0014 | — | 100 | 0 | 0 |