Solves the problem of 'co-firing' conflicts in probabilistic ML routing systems using temperature-scaled softmax partitioning.
arXiv · March 20, 2026 · 2603.18174
The Takeaway
In multi-agent or multi-model routing (e.g., Semantic Router), independent thresholds often lead to a single query triggering multiple conflicting rules. This framework uses Voronoi partitioning to ensure disjoint routing without model retraining, providing a mathematically sound way to build complex agentic pipelines.
From the abstract
Conflict detection in policy languages is a solved problem -- as long as every rule condition is a crisp Boolean predicate. BDDs, SMT solvers, and NetKAT all exploit that assumption. But a growing class of routing and access-control systems base their decisions on probabilistic ML signals: embedding similarities, domain classifiers, complexity estimators. Two such signals, declared over categories the author intended to be disjoint, can both clear their thresholds on the same query and silently