Data Dependency Inference for Industrial Code Generation Based on UML Sequence Diagrams

Wenxin Mao, Zhitao Wang, Long Wang, Sirong Chen, Cuiyun Gao, Luyang Cao, Ziming Liu, Qiming Zhang, Jun Zhou, Zhi Jin, {lucascao, simonzmliu, archrzhang, anderszhou}@tencent.com, [email protected] · 2025

Read Data Dependency Inference for Industrial Code Generation Based on UML Sequence Diagrams by Wenxin Mao, Zhitao Wang, Long Wang, Sirong Chen, Cuiyun Gao, Luyang Cao, Ziming Liu, Qiming Zhang, Jun Zhou, Zhi Jin, {lucascao, simonzmliu, archrzhang, anderszhou}@tencent.com, [email protected] (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-2508-03379.txt.

Description

Wenxin Mao and co-authors teach language models to write factory software from diagrams rather than vague prose. Plain text requirements hide tangled behaviors and data flows. That is common in service-based systems. Their framework starts from a rich sequence diagram. It adds decision tables and interface specs, which removes guesswork from the wording.

Before any code is written, the system first works out a clear graph of data links between services. They frame that step as a math-style reasoning task with tight rules. That suits what the models do best. Static checks then prune the graph to keep it manageable. The authors argue that making the links clear first should raise how often the code comes out right.

Catalog record

CREATORWenxin Mao, Zhitao Wang, Long Wang, Sirong Chen, Cuiyun Gao, Luyang Cao, Ziming Liu, Qiming Zhang, Jun Zhou, Zhi Jin, {lucascao, simonzmliu, archrzhang, anderszhou}@tencent.com, [email protected]
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE51 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2508-03379

Browse the full library · About the Tower of Babel