Tag
The EchoNet benchmark tests how open-weight AI models handle fake sources during agentic search, showing that models struggle most against fake majorities and revealing cost and performance differences.