TAM-Eval and RM-RF from RnD Center T-Technologies (Category Engineering)

Our R&D center presented pair TAM-Eval and RM-RFIt makes the generation of modular tests using LLM noticeably more practical for visual design.

TAM-Eval sets reproducible standard test quality scores in live repositories, and RM-RF can predict which tests are actually useful before assembly and launch. In practice, this means smaller runways, less strain on infrastructure and a CI/CD cycle. This is especially important in a world that is rapidly moving towards AI-native development. (see. posture) : when code and tests increasingly generate models, the main competitive advantage is not only generation, but also reliable, fast quality assessment. . Without this, AI acceleration easily turns into AI chaos.

These articles are well received by the academic community: RM-RF is accepted in mainstay SANER 2026 and TAM-Eval on VST 2026.

Finally, I will quote Stanislav Moiseev, head of the RnD center, who described the benefits of these methods in this way.

These methods make the operation of large language models with tests more predictable and efficient for real-world development processes. TAM-Eval sets the standard for comparing models and agents, accompanied by tests on measurable metrics, and RM-RF allows you to weed out weak tests and rank strong ones without expensively running a piplin at every step.

#Engineering #RnD #AI #DevOps #Software