Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Liu, Xunzhuo, Wu, Hao, Chen, Huamin, He, Bowei, Liu, Xue
Format: Preprint
Veröffentlicht: 2026
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2603.18174
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866914408528084992
author Liu, Xunzhuo
Wu, Hao
Chen, Huamin
He, Bowei
Liu, Xue
author_facet Liu, Xunzhuo
Wu, Hao
Chen, Huamin
He, Bowei
Liu, Xue
contents 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 route it to the wrong model. Nothing in the compiler warns about this. We characterize the problem as a three-level decidability hierarchy -- crisp conflicts are decidable via SAT, embedding conflicts reduce to spherical cap intersection, and classifier conflicts are undecidable without distributional knowledge -- and show that for the embedding case, which dominates in practice, replacing independent thresholding with a temperature-scaled softmax partitions the embedding space into Voronoi regions where co-firing is impossible. No model retraining is needed. We implement the detection and prevention mechanisms in the Semantic Router DSL, a production routing language for LLM inference, and discuss how the same ideas apply to semantic RBAC and API gateway policy.
format Preprint
id arxiv_https___arxiv_org_abs_2603_18174
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Conflict-Free Policy Languages for Probabilistic ML Predicates: A Framework and Case Study with the Semantic Router DSL
Liu, Xunzhuo
Wu, Hao
Chen, Huamin
He, Bowei
Liu, Xue
Machine Learning
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 route it to the wrong model. Nothing in the compiler warns about this. We characterize the problem as a three-level decidability hierarchy -- crisp conflicts are decidable via SAT, embedding conflicts reduce to spherical cap intersection, and classifier conflicts are undecidable without distributional knowledge -- and show that for the embedding case, which dominates in practice, replacing independent thresholding with a temperature-scaled softmax partitions the embedding space into Voronoi regions where co-firing is impossible. No model retraining is needed. We implement the detection and prevention mechanisms in the Semantic Router DSL, a production routing language for LLM inference, and discuss how the same ideas apply to semantic RBAC and API gateway policy.
title Conflict-Free Policy Languages for Probabilistic ML Predicates: A Framework and Case Study with the Semantic Router DSL
topic Machine Learning
url https://arxiv.org/abs/2603.18174