SOURCE-LINKED INTELLIGENCE

Selection-Aware Stress Testing for Interactive Agents

arXiv · AI, language, vision and robotics · article · Aug 31, 2026 · UTC

Agent evaluations often use one benchmark to choose a workflow and then search for task types where its advantage weakens, so both conclusions are selected from the same data. We introduce Selection-Aware Semantic Stress Testing (\SASST{}), which learns a task reweighting from pre-execution features on discovery tasks and evaluates the same paired comparison on separate confirmation tasks. The protocol checks support and stability, uses joint bounds for all planned claims, and can return no claim. We prove conditional asymptotic validity under stated cluster assumptions. A forty-cluster audit

Read original source ↗ Open in workspace

recordType

paper

region

Global