Retrieval-Augmented TLAPS Proof Generation with Large Language Models

Yuhao Zhou · 2025

Read Retrieval-Augmented TLAPS Proof Generation with Large Language Models by Yuhao Zhou (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-2501-03073.txt.

Description

Formal proofs can be drafted by language models with a library at their elbow, in a 2025 paper by Yuhao Zhou. The target is TLAPS, the proof engine for the TLA+ specification language used to verify critical systems.

The method works in two passes. First it breaks a hard proof obligation into smaller sub-obligations. Then it writes each proof step while retrieving verified examples of similar proofs for guidance. Tests ranged from basic arithmetic facts to properties of algorithms. The system produced valid proofs for middling obligations but stalled on the hardest theorems. Zhou reads the result as a practical assistant for proof engineers, strongest exactly where routine verification work piles up.

Catalog record

CREATORYuhao Zhou
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE57 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2501-03073

Browse the full library · About the Tower of Babel