Does the groundedness classifier check whether the context itself is correct?

No. It checks whether the answer matches the context the agent was given, not whether that context is accurate, so an answer built faithfully on a wrong or stale source still scores grounded.

Groundedness answers one narrow question: did the agent stay inside what it was handed, or did it add something the context never said. If a tool returns a stale price and the agent repeats that price exactly, the answer is grounded, it’s the input that was wrong. That’s a different failure, upstream of anything this classifier is built to catch.

Keeping the check that narrow is what makes it cheap enough to run on every trace instead of a sample. A classifier that also had to judge the correctness of every document, tool result, and prior turn it saw would be a much harder, slower problem, and a different one from noticing when an agent invents something its own context doesn’t support.

keep reading

More on this.

Can a factually correct answer still get flagged as ungrounded?Yes. Groundedness checks whether an answer is supported by its context, not whether the answer is true, and those are different questions.answer →Does groundedness check every LLM call, or just the final answer?Every LLM call. It scores each call against its own context as the call happens, not just the last message a trace sends back to the user.answer →Does the groundedness classifier only work on agents that use RAG?No. It scores any LLM call against whatever context that call received, tool outputs and prior turns included, not only retrieved documents.answer →Everything on tessary groundednessThe definition, and every question answered under tessary groundedness.concept →

Send us the traces you already emit.

book a demo