An AI system makes software tests that better match developers’ intended checks

The system uses a developer’s description of a test scenario and expected result to create a project-specific test.

Editorial illustration — not from the study.

Most automated test-generation tools aim to cover more branches of a program or turn a particular piece of code into a test. IntentionTest instead starts with a validation intention: the test scenario, precondition and expected result a developer wants to check.

The system finds a reusable test from the same project, then uses a large language model to edit it for the new intention. It also supplies relevant project code and global variables as context, helping the generated test use the project’s existing patterns.

What the system produced

In an evaluation of 3,680 test cases from 12 open-source projects, IntentionTest was compared with four other methods: TELPA, DA, ChatTester and EvoSuite. Given a validation intention, it generated tests that were more semantically relevant to the reference tests. Its common mutation scores were 28.1% to 37.6% higher, common coverage ratios were 16.9% to 23.9% higher, and successful-pass rates were 23.7% to 49.0% higher than those of the comparison methods.

Why intended checks matter

A test can cover program code without checking the behavior a developer actually cares about. By using the intended scenario and expected result, along with reusable tests and project-specific code details, IntentionTest addresses information that branch coverage and code-to-test translation do not directly provide. The results suggest this approach can produce tests that are more closely aligned with a project’s existing testing practices.

Evidence and caveats

The evidence comes from a comparative evaluation against four baselines on 3,680 test cases from 12 open-source projects. The findings therefore show how IntentionTest performed in those tested projects and scenarios; the abstract does not establish how it would perform across all software projects or intentions. The evaluation also depends on having a written validation intention, and the abstract does not report whether developers adopted or maintained the generated tests in practice.

Programmable hardware speeds matrix calculations up to 6.3 times

Researchers tested a programmable hardware design that imitates selected quantum-computing ideas to run matrix calculations in parallel. In representative benchmarks, the design reached up to 6.3 times the speed and 4.1 times the energy efficiency of conventional matrix co-processors and embedded graphics processors.

2 days ago·13 views

AI & Computing

AI and docking agreed in some drug screens, but not all

Researchers evaluated three AI systems on hundreds of molecules bound to a SARS-CoV-2 protein and in three prospective drug screens. The results suggest that combining AI-based predictions with traditional docking may help prioritize candidates, although no single approach worked best in every test.

4 days ago·50 views

AI & Computing

A hardware monitor detected simulated attacks on car networks in 14 microseconds

Researchers developed a hardware-assisted system intended to detect message injection, spoofing, flooding and impersonation attacks on automotive networks. In tests using simulated traffic and a prototype automotive processor system, it reported an average detection time of 14 microseconds and less than 2 percent CPU overhead.

5 days ago·42 views