Нейроразбор резюме: практический опыт и нюансы создания LLM-судей — Женя Орлов, hh⁠.ru

HR tech companies already use LLM to evaluate candidates, but they regularly face a challenge: such assessments are difficult to verify, standardize, and make reproducible. To understand how reliable neural models can be in this task, we at hh developed an approach using an LLM judge—a separate model that evaluates the quality of responses and resumes based on specified criteria.