A Gradual Probabilistic Lambda Calculus
summary
Defines GPLC, a gradual probabilistic lambda calculus with probabilistic types and a binary choice operator, and elaborates programs to a target TPLC with distribution semantics. Types use couplings and generate SMT-style constraints over symbolic probabilities; satisfiability yields ranges (e.g. choose choice probability ≤ 37.5% to hit a 95% uptime). Proves type safety, conservative extension of the static calculus, and the gradual guarantee using coupling-based consistency constraints checked at runtime.
AI-written summary · about