Conflict-Free Policy Languages for Probabilistic ML Predicates A Framework and Case Study with the Semantic Router DSL
Xunzhuo Liu, Hao Wu, Huamin Chen, [4pt] Bowei He, Xue Liu, [6pt] vLLM Semantic Router Project, [1pt] MBZUAI / McGill University · 2026
Read Conflict-Free Policy Languages for Probabilistic ML Predicates A Framework and Case Study with the Semantic Router DSL by Xunzhuo Liu, Hao Wu, Huamin Chen, [4pt] Bowei He, Xue Liu, [6pt] vLLM Semantic Router Project, [1pt] MBZUAI / McGill University (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-18174.txt.
Description
Liu and co-authors build conflict-free policy languages for systems driven by probabilistic ML signals. Crisp rule conflicts are decidable, but embedding and classifier signals can both fire on one query. They map a three-level hierarchy: SAT for crisp rules, spherical caps for embeddings, and undecidable classifier cases. A temperature-scaled softmax splits embedding space into Voronoi regions where co-firing cannot occur. The tools ship in the Semantic Router DSL for LLM routing.
Catalog record
| CREATOR | Xunzhuo Liu, Hao Wu, Huamin Chen, [4pt] Bowei He, Xue Liu, [6pt] vLLM Semantic Router Project, [1pt] MBZUAI / McGill University |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 31092 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2603.18174