BALI : Branch-Aware Loop Invariant Inference with Large Language Models

Mingxiu Wang, Jiawei Wang, Xiao Cheng · 2026

Read BALI : Branch-Aware Loop Invariant Inference with Large Language Models by Mingxiu Wang, Jiawei Wang, Xiao Cheng (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-2601-00882.txt.

Description

Mingxiu Wang and two co-authors present BALI in 2026, a tool that finds loop invariants with help from AI models. A loop invariant is a fact that stays true each time a program loop runs. Finding them by hand is slow, and hard for tangled code. AI models can guess them, but plain guess-and-check wastes effort.

BALI first splits a program by the paths its branches can take. For each path, it asks the AI for clause-level facts and checks them with a logic solver. True clauses are then joined into a full invariant for the loop. That order, check small then join, is the core twist. The authors lay out the parts of the system and give early results. They also map what is left to build. The goal is a fully automatic route to proofs of code.

Catalog record

CREATORMingxiu Wang, Jiawei Wang, Xiao Cheng
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE19 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2601-00882

Browse the full library · About the Tower of Babel