Measuring hallucinations in document Q&A: methods that hold up
How to measure hallucinations in a document Q&A or RAG system: claim-level faithfulness, citation checks, published benchmarks and hand grading.
Read articleCategory
AI Automation
Articles about ai automation, with links to the sources behind each claim.
RSS feed5 articles
Human-in-the-loop design: where approval steps belong in AI workflows
Where to put human approval in an AI workflow, what the reviewer should see, and how to stop approvals turning into a rubber stamp.
Read articleWhen not to use an LLM: tasks where plain code and rules win
How to tell which steps in a workflow need a language model and which are better as plain code or rules, with a simple test and a hybrid pattern.
Read articleStructured outputs and function calling for reliable LLM automation
What structured outputs and function calling guarantee, where the guarantee stops, and how to design schemas and checks so an LLM can safely feed an automation.
Read articleHow to evaluate an AI automation before you trust it
Build a test set from real work, grade outputs with code and careful LLM judges, read the failures, and decide where human review stays before you go live.
Read article