בעברית, עד 90 תווים: Absorbed in Inertia: Activation Analysis for Computer-Use Agents

תקציר מקורי באנגליתarXiv:2609.37176v1 Announce Type: new Abstract: Computer-use agents have become increasingly capable of executing tasks on live desktops through natural-language instructions, based on trajectories of screenshots, actions, and reasoning. We discover that they can stealthily exhibit inertia, in which they repeat fruitless actions despite recognizing that these actions are ineffective. We hypothesize that inertia is reflected in the agent's internal state, i.e., the activation values of the agent's underlying model, and propose a protocol to measure the relationship between the two. Extensive analysis of high-dimensional activation states shows that inertia corresponds to an absorbing region of activation space, where activation values become stale across actions and even after attempts to s