https://blog.lameiro0x.com/tags/evals/Recent content in Evals on Miguel Lameiro | Cybersecurity Blog & Security WriteupsHugo -- 0.161.1en-usTue, 05 May 2026 00:00:00 +0000https://blog.lameiro0x.com/notes/ai-security/automated-evals-for-llmops-testing-llm-apps-in-ci/Tue, 05 May 2026 00:00:00 +0000https://blog.lameiro0x.com/notes/ai-security/automated-evals-for-llmops-testing-llm-apps-in-ci/<h2 id="introduction">Introduction</h2> <p>Traditional software tests usually compare a known input with a predictable output. LLM applications are different because the output is generated, variable, and sometimes correct in more than one form.</p> <p>That does not mean LLM apps cannot be tested. It means the test suite needs several layers: deterministic checks where possible, model-graded evaluations where judgment is required, hallucination checks against known context, and CI automation so regressions are caught before release.</p>