Search papers, labs, and topics across Lattice.
The paper introduces EntroRouter, a novel single-round model routing framework that addresses the issue of Trust Region Collapse, a failure mode where capable models are suppressed due to the coupling of reasoning and routing. By employing entropy regulation as a core objective, the approach initializes a high-entropy prior for exploration through Soft Supervision and stabilizes the learning process with a Soft Anchor that manages entropy contraction. Experimental results show that EntroRouter achieves 98.3% accuracy of the best expert while significantly reducing computational costs by 48.25%, highlighting its efficiency and effectiveness in model routing.
EntroRouter achieves near-optimal expert accuracy while slashing computational costs by nearly half, challenging the status quo of model routing efficiency.
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%.
