Cost-Driven Synthesis of Sound Abstract Interpreters

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Gu, Qiuhan, Singh, Avaljot, Singh, Gagandeep
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909908909162496
author Gu, Qiuhan
Singh, Avaljot
Singh, Gagandeep
author_facet Gu, Qiuhan
Singh, Avaljot
Singh, Gagandeep
contents Constructing abstract interpreters that provide global soundness guarantees remains a major obstacle in abstract interpretation. We investigate whether modern LLMs can reduce this burden by leveraging them to synthesize sound, non-trivial abstract interpreters across multiple abstract domains in the setting of neural network verification. We formulate synthesis as a constrained optimization problem and introduce a novel mathematically grounded cost function for measuring unsoundness under strict syntactic and semantic constraints. Based on this formulation, we develop a unified framework that unifies LLM-based generation with syntactic and semantic validation and a quantitative cost-guided feedback mechanism. Empirical results demonstrate that our framework not only matches the quality of handcrafted transformers, but more importantly, discovers sound, high-precision transformers for complex nonlinear operators that are absent from existing literature.
format Preprint
id arxiv_https___arxiv_org_abs_2511_13663
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Cost-Driven Synthesis of Sound Abstract Interpreters
Gu, Qiuhan
Singh, Avaljot
Singh, Gagandeep
Programming Languages
Machine Learning
Constructing abstract interpreters that provide global soundness guarantees remains a major obstacle in abstract interpretation. We investigate whether modern LLMs can reduce this burden by leveraging them to synthesize sound, non-trivial abstract interpreters across multiple abstract domains in the setting of neural network verification. We formulate synthesis as a constrained optimization problem and introduce a novel mathematically grounded cost function for measuring unsoundness under strict syntactic and semantic constraints. Based on this formulation, we develop a unified framework that unifies LLM-based generation with syntactic and semantic validation and a quantitative cost-guided feedback mechanism. Empirical results demonstrate that our framework not only matches the quality of handcrafted transformers, but more importantly, discovers sound, high-precision transformers for complex nonlinear operators that are absent from existing literature.
title Cost-Driven Synthesis of Sound Abstract Interpreters
topic Programming Languages
Machine Learning
url https://arxiv.org/abs/2511.13663