AI TrackerFrontier desk · AEST
Back to news
normalaitech

VeriCUA: Verification as Agentic Exploration for Computer-Using Agents

arXiv cs.AI · 30 Oct 2026, 03:45 AEST

VeriCUA is a platform-agnostic framework designed for verification through agentic exploration in computer-using agents. It aims to improve task-completion verification across various environments by dynamically gathering and assessing evidence.

Key facts

Read aloud

VeriCUA is a platform-agnostic framework designed for verification through agentic exploration in computer-using agents. It aims to improve task-completion verification across various environments by dynamically gathering and assessing evidence. Building and scaling computer-using agents CUAs across desktop, mobile, and web environments requires reliable task-completion verification for data curation, evaluation, and reinforcement learning. Typical automated approaches prompt vision-language models VLMs to judge trajectories or generate programs to inspect application state. We present VeriCUA, a platform-agnostic framework for verification through agentic exploration. Experiments show improved agreement with reference labels over the compared baselines across these settings, with the largest offline gains on hard cases and stronger rejection of false completion claims online.

Sources in this cluster