Sakura: From Natural-Language Descriptions to Complex Tests
Sakura explores how to generate complex software tests from abstract natural-language descriptions, making high-level testing intent more actionable.
This research note is currently being developed. It will provide a concise, accessible walkthrough of the motivation behind Sakura, the approach we explored, and what we learned from evaluating it.
The problem
Natural-language test descriptions can express useful testing intent without prescribing every implementation detail. Sakura investigates how those abstract descriptions can guide the generation of complex, executable tests.
What this note will cover
The completed note will explain the core idea behind Sakura, walk through an example, summarize the main results, and discuss what the work suggests for future test-generation systems. Until then, the paper and source code are available above.