Predictive LLM Routing on AlpacaEval
63.17Score (vs OpenAI)Oracle
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Oracle2025.05 | 63.17 | 52.98 | 34.69 | |
| Linear2025.05 | 57.6 | 52.84 | 34.26 | |
| MLPRepresentation=Matrix Factorization (MF)2025.05 | 57.5 | 52.84 | 34.2 | |
| kNNk=1002025.05 | 57.38 | 52.77 | 34.26 | |
| Graphk=1002025.05 | 57.37 | 51.84 | 32.25 | |
| kNNk=102025.05 | 57.33 | 52.82 | 34.27 | |
| D-Attnk=1002025.05 | 57.17 | 52.77 | 24.45 | |
| LinearRepresentation=Matrix Factorization (MF)2025.05 | 56.85 | 52.08 | 33.76 | |
| Graphk=102025.05 | 55.66 | 45.64 | 31.5 | |
| MLP2025.05 | 55.5 | 51.29 | 32.18 | |
| Attnk=1002025.05 | 55.29 | 52.49 | 27.94 | |
| MLPMatrix Factorization=true2025.05 | 54.71 | 51.14 | 30.96 | |
| Linear2025.05 | 54.61 | 51.14 | 30.96 | |
| Graphk=1002025.05 | 54.61 | 51.14 | 30.96 | |
| LinearMatrix Factorization=true2025.05 | 54.54 | 51.14 | 30.96 | |
| Graphk=102025.05 | 54.52 | 51.14 | 30.95 | |
| D-Attnk=102025.05 | 54.5 | 44.62 | 29.34 | |
| Attnk=102025.05 | 54.43 | 45.12 | 26.52 | |
| Attnk=102025.05 | 54.08 | 51.14 | 30.91 | |
| Attnk=1002025.05 | 54.04 | 51.14 | 30.96 | |
| D-Attnk=102025.05 | 53.84 | 51.14 | 30.91 | |
| D-Attnk=1002025.05 | 53.55 | 51.14 | 30.96 | |
| MLP2025.05 | 50.81 | 51.14 | 30.52 | |
| kNNk=102025.05 | 38.32 | 42.04 | 23.72 | |
| kNNk=1002025.05 | 37.89 | 43.99 | 23.88 | |
| Random2025.05 | 36.47 | 34.56 | 27.76 |