List your company

Agentic Model Routing on In-Distribution n=2,919 (test)

57Average Performance (%)Oracle
35.846441.338246.8352.3218Jun 22, 2026

Evaluation results

MethodLinks
OracleTaxonomy=Oracle2026.065708.2
ACRouterTaxonomy=Agent-as-a-Router2026.0649.98205.53.79
DimensionBestTaxonomy=Static: Heuristic2026.0647.5277.43.69
LogRegTaxonomy=Static: Trained Policy2026.0647.26284.46.27
RouteLLM-BERTTaxonomy=Static: Trained Policy2026.0647.22285.56.22
kNN RetrievalTaxonomy=Static: Heuristic2026.0647.18286.76.07
TF-IDF+MLPTaxonomy=Static: Trained Policy2026.0646.97292.86.11
LinUCBTaxonomy=Dynamic: Online Bandit2026.0646.84296.94.38
LinTSTaxonomy=Dynamic: Online Bandit2026.0646.48307.44.49
Qwen3.5-0.8B-FinetunedTaxonomy=Static: Trained Policy2026.0646.41309.16.82
RouteLLM-MFTaxonomy=Static: Trained Policy2026.0646.16316.56.19
Always-Opus 4.6Taxonomy=Single-Model Baselines2026.0643.83387.11.29
RandomTaxonomy=Single-Model Baselines2026.0638.75533.62.48
Always-Qwen3.5-PlusTaxonomy=Single-Model Baselines2026.0637.16580.22.05
Always-Kimi-K2.5Taxonomy=Single-Model Baselines2026.0636.66593.312.62