Abstract
A system for automated code verification can address challenges associated with verifying high-velocity code changes, where manual review may be slow and broad automated testing might lack specificity. The system can utilize a multi-agent approach for differential trace analysis. A synthesizer agent can, for example, analyze a specific code change to generate a targeted test plan. This plan may be executed in parallel on a control version of a software component without the change and an experiment version that includes it. During execution, observation agents can capture detailed execution traces and performance metrics from both runs. The system may first verify that the test path exercised the modified code before performing a differential analysis on the traces to detect potential semantic or performance regressions. This process can provide a targeted, automated indication of correctness for software changes within a development pipeline.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Cheng, Shu-Wei; Wang, Fei; Tsai, Yao-Sheng; IMMADISETTY, AKHILA; and Li, Zening, "Automated Code Change Verification via Differential Trace Analysis", Technical Disclosure Commons, (September 02, 2026)
https://www.tdcommons.org/dpubs_series/11576