Multi-Language Probabilistic Programming

Sam Stites, John M. Li, Steven Holtzen · 2025

Read Multi-Language Probabilistic Programming by Sam Stites, John M. Li, Steven Holtzen (2025) 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-2502-19538.txt.

Introduction

Sam Stites and co-authors argue that users of probabilistic programming should be able to mix languages mid-task. Each such language shines at one kind of model, yet users must pick one up front and stay locked in. Real problems often blend kinds, so the lock-in hurts.

Their system, MultiPPL, joins two languages with opposite strengths. One does exact reasoning on discrete models at high speed. The other uses importance sampling for models exact tools cannot touch. The paper gives syntax and meaning rules for the blend. It proves the joint inference method sound, so mixed programs still yield valid answers. Tests on mixed models show the blend exploits each side where it is strong. The work lays a base for freer mixing across more languages.

Catalog record

CREATORSam Stites, John M. Li, Steven Holtzen
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE265 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2502-19538

Browse the full library · About the Tower of Babel