{"slug": "google-and-university-of-hong-kong-researchers-want-ai-agents-to-show-their-work", "title": "Google and University of Hong Kong researchers want AI agents to show their work", "summary": "A research paper from Google and the University of Hong Kong introduces VeriCUA, a verifier that inspects app and file state rather than trusting a computer-using agent's claim that a task is complete. On the OSWorld benchmark VeriCUA cut the false-positive rate from 27.9% to 13.9%, on ScienceBoard from 24.8% to 5.4%, and on AndroidWorld it reached 90.5% accuracy with a 7.7% false-positive rate versus 82.4% accuracy and 16.9% for an auto-generated script verifier, with decision accuracy above 95% when the system reports confidence. The paper, listed on arXiv as 2610.40578 and submitted October 29, 2026, with authors including Qiushi Sun and Pritish Kamath, reports that in online tests VeriCUA rejected approximately 80% of false completion claims.", "body_md": "# Google and University of Hong Kong researchers want AI agents to show their work\n\nGoogle / Wikimedia Commons (Public domain)\n\nA new paper introduces VeriCUA, a verifier that inspects app data instead of trusting an AI agent's claim that a task is done\n\nAI agents have a habit familiar to anyone who has managed a summer intern. They say the job is done, and sometimes it isn’t.\n\nA new research paper from [Google](https://cryptobriefing.com/markets/alphabet/) and the University of Hong Kong takes direct aim at that problem. It introduces **VeriCUA**, a framework that checks the actual state of apps and files before accepting an agent’s claim that a task is complete.\n\nThe core idea is simple. An agent saying “finished” counts as a claim, not as proof.\n\n## What VeriCUA actually does\n\nThe paper focuses on computer-using agents, or CUAs. These are AI systems that operate software the way a person would, by clicking through interfaces, typing, and moving between apps.\n\nVeriCUA takes what the authors describe as an agentic approach to verification. It first builds a rubric tailored to the specific task, drawing on the task’s context.\n\nThen it goes looking for evidence. The verifier runs GUI-based or code-based checks, such as reading files and inspecting system state, before it issues a verdict.\n\nThe paper, listed on arXiv as 2610.40578, was submitted on October 29, 2026. Its authors include Qiushi Sun and Pritish Kamath.\n\n## The numbers behind the pitch\n\nThe researchers tested VeriCUA across desktop, mobile, and web environments. According to the paper, it beat baseline systems across the board, with the biggest gains showing up in harder scenarios.\n\nThe headline metric is the false-positive rate. In this context, a false positive means the verifier signs off on a task the agent did not actually complete.\n\n### AI, tech, and the markets they move—in one daily briefing.\n\nDaily. Free. Join 34,000+ readers across crypto, finance, and policy.\n\nOn the OSWorld benchmark, VeriCUA cut that rate from 27.9% to 13.9%. That is roughly half the bogus approvals of the comparison method.\n\nThe ScienceBoard results were more dramatic. The false-positive rate fell from 24.8% to 5.4%.\n\nOn AndroidWorld, a mobile benchmark, VeriCUA posted 90.5% accuracy with a false-positive rate of 7.7%. The baseline, an auto-generated script verifier, recorded 82.4% accuracy and a 16.9% false-positive rate.\n\nThe paper also reports that when VeriCUA is confident in its judgment, decision accuracy exceeds 95%. The system also showed higher agreement with reference labels than the baselines did.\n\nIn online tests, VeriCUA rejected approximately 80% of false completion claims.\n\n## Why self-reports are a problem\n\nThe broader issue the paper raises is inflated success rates. When verification relies on weak checks, agents can appear more capable than they really are.\n\nThe researchers position their dynamic verification approach as a counter to that inflation. If the checker only looks at surface signals, a confident-sounding agent can pass tests it should fail.\n\nThe paper drew discussion in the tech community shortly after it appeared on arXiv.\n\n## What this means for AI agents\n\nFor teams building computer-using agents, the takeaway is uncomfortable but useful. If older verification methods let through false positives at rates like 27.9% or 24.8%, some reported success numbers may be flattering the underlying systems.\n\nThe approximately 80% rejection rate for false completion claims in online testing suggests that layer can catch a meaningful share of those quiet failures.\n\nThe confidence signal is also notable. Decision accuracy above 95% when the system is confident means builders could potentially route uncertain cases to humans and let the verifier handle the clearer ones.\n\nFor now, the research makes a straightforward argument. When an AI agent says it finished the job, somebody should check. Google and the University of Hong Kong are proposing that the somebody be another agent, one that actually opens the files.\n\n**Disclosure:** This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our [Editorial Policy](https://cryptobriefing.com/editorial-policy/).\n\n## More in AI\n\n[See all](https://cryptobriefing.com/artificial-intelligence/)", "url": "https://wpnews.pro/news/google-and-university-of-hong-kong-researchers-want-ai-agents-to-show-their-work", "canonical_source": "https://cryptobriefing.com/google-hku-vericua-ai-agent-verification/", "published_at": "2026-11-03 02:41:24+00:00", "updated_at": "2026-11-03 02:50:29.373242+00:00", "lang": "en", "topics": ["ai-agents", "ai-research", "artificial-intelligence", "ai-safety"], "entities": ["Google", "University of Hong Kong", "VeriCUA", "Qiushi Sun", "Pritish Kamath", "OSWorld", "ScienceBoard", "AndroidWorld"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/google-and-university-of-hong-kong-researchers-want-ai-agents-to-show-their-work", "markdown": "https://wpnews.pro/news/google-and-university-of-hong-kong-researchers-want-ai-agents-to-show-their-work.md", "text": "https://wpnews.pro/news/google-and-university-of-hong-kong-researchers-want-ai-agents-to-show-their-work.txt", "jsonld": "https://wpnews.pro/news/google-and-university-of-hong-kong-researchers-want-ai-agents-to-show-their-work.jsonld"}}