Multi-agent routing on WildChat-12 (test)
97.8PrecisionEncoder
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| EncoderThreshold=0.602026.06 | 97.8 | 85.98 | 89.59 | 85.52 | 73.11 | 1.2 | |
| MLThreshold=0.602026.06 | 73.03 | 75.33 | 71.79 | 66.36 | 49.67 | 1.5 | |
| MLkNNThreshold=0.602026.06 | 71.44 | 60.5 | 63.1 | 58.36 | 45.33 | 1.15 | |
| CCThreshold=0.602026.06 | 64.49 | 71.96 | 65.13 | 58 | 37.56 | 1.65 | |
| KNNThreshold=0.602026.06 | 52 | 38.39 | 42.56 | 38.39 | 27 | 1 | |
| LLMThreshold=0.60, Zero-shot=true2026.06 | 46.22 | 42.33 | 41.51 | 35.56 | 19 | 1.4 | |
| MajorityThreshold=0.602026.06 | 17.33 | 12.33 | 13.89 | 12.33 | 8 | 1 |