EntroRouter: Learning Efficient Model Routing via Entropy Regulation
For practitioners deploying multiple models, EntroRouter offers a practical routing method that balances accuracy and cost, though the improvement is incremental over existing routing frameworks.
EntroRouter addresses Trust Region Collapse in model routing by decoupling reasoning and routing via entropy regulation, achieving 98.3% of the strongest expert's accuracy with 48.25% cost reduction.
Model routing balances solution accuracy and computational cost by selecting among models of varying capabilities. While recent multi-round frameworks interleave reasoning and planning, we identify a structural failure mode termed Trust Region Collapse. We demonstrate that the deep coupling of reasoning and routing, exacerbated by the dominance of strong pre-training priors under sparse supervision, leads to degenerate local optima where capable experts are systematically suppressed. To decouple these processes, we propose $\textbf{EntroRouter}$, a single-round routing framework that treats entropy regulation as a core objective. We first initialize the policy via Soft Supervision, fitting a distribution of suitable models to establish a high-entropy prior for exploration. Subsequently, we stabilize Reinforcement Learning using a Soft Anchor, which utilizes offline capability estimates to orchestrate controlled entropy contraction within a safe trust region. Extensive experiments demonstrate that EntroRouter retains 98.3% of the strongest expert's accuracy while reducing computational costs by 48.25%.