Terry Migwi·1h agoBuilding a Search System, Part 4: RAG EvaluationThis article documents how I evaluated the hybrid search pipeline against a structured dataset of 12 questions using two metrics, accuracy…
Zelda Ailine Luconi·8h agoGoogle ADK 2.0 Tutorial (#4b): Running Evaluations with adk evalHands-on guide to running, debugging, and automating multi-turn agent evaluations with Google Cloud and ADK 2.0. Covers evaluation…
Amrutalohabare·20h agoTwo Tests Failed. Only One Was the AI’s Fault.Learning to test AI, Session 7 (your first automated eval with promptfoo): 15 tests, two models, one command, and why the score was the…
InProvectus ArticlesbyProvectus·1d agoHow We Test Barley’s Answers: Evaluation with DeepEval and LangSmithIn our previous articles we talked a lot about Barley — our AI-powered virtual assistant and his features but there is one aspect we…
Ali Nasir·1d agoYour AI Model Passed the Benchmark. Is It Ready for Production?Why enterprise AI needs release gates that test business outcomes, security, reliability, and cost — not just model accuracy.
Adam Chang·1d agoWhen Your AI Agent Fails a Test, Check the Judge First: Evaluating a Refund Agent With 7 DeepEval…Several of the test failures in my refund agent's evaluation suite weren't the agent's fault. The LLM judge couldn't see what the customer…A response icon2A response icon2
Serhii Smetanskyi·2d agoTesting LLMs with Jev-Based Evals: Clef as the JudgeLLM-as-judge has three problems that show up as soon as the suite grows. The first is cost. Every test case is a paid call to a large…A response icon1A response icon1
InTowards AIbyAlex Punnen·3d agoHow to Build an AI Agent Eval Framework With Open Source ToolsOne contract file, three layers of evals, and the bad answer that got past my keyword checkA response icon1A response icon1
Gaurav Goyal·3d agoThree ways an AI system’s cited numbers went wrong, and how to catch themEvery answer had citations, and the eval scored most of them as well grounded. When I checked each number against the excerpt it cited…A response icon1A response icon1
