List your company

Agentic Model Routing on OOD n=176 (test)

75.89Average Performance (%)Oracle
-0.248419.518339.28559.0517Jun 22, 2026

Evaluation results

MethodLinks
OracleTaxonomy=Oracle2026.0675.8902.32
ACRouterTaxonomy=Agent-as-a-Router2026.0662.5171.18
Always-Opus 4.6Taxonomy=Single-Model Baselines2026.0657.1426.70.64
Qwen3.5-0.8B-FinetunedTaxonomy=Static: Trained Policy2026.0655.3627.20.74
LinUCBTaxonomy=Dynamic: Online Bandit2026.0649.8231.10.96
LinTSTaxonomy=Dynamic: Online Bandit2026.0646.4335.90.75
RandomTaxonomy=Single-Model Baselines2026.0631.2550.40.85
RouteLLM-BERTTaxonomy=Static: Trained Policy2026.0621.4359.41.3
LogRegTaxonomy=Static: Trained Policy2026.0619.6461.81.17
Always-Kimi-K2.5Taxonomy=Single-Model Baselines2026.0618.7562.31.22
kNN RetrievalTaxonomy=Static: Heuristic2026.0614.2966.71.45
TF-IDF+MLPTaxonomy=Static: Trained Policy2026.0613.3967.91.17
RouteLLM-MFTaxonomy=Static: Trained Policy2026.068.9372.70.94
Always-Qwen3.5-PlusTaxonomy=Single-Model Baselines2026.062.6880.10.19