In the news
Evaluating VLMs for Autonomous Agent-Driven Geometry Clipping Detection in Video Game QA
arXiv cs.AI · Published · 3 min read
In 30 seconds
- What happened
- Researchers evaluated six Vision-Language Models for detecting geometry clipping bugs in video games using autonomous agents and zero-shot prompting.
- Why it matters
- Game QA engineers considering automated visual bug detection should understand VLM capabilities and limitations for this specific anomaly detection task.
- Watch out
- All tested VLMs produced substantial false positives on ambiguous frames like near-contact geometry and partial occlusions, limiting standalone use.
Listen to this summary
From the article
-->
Computer Science > Computer Vision and Pattern Recognition
arXiv:2607.25921v1 (cs)
[Submitted on 28 Jul 2026]
Title: Evaluating VLMs for Autonomous Agent-Driven Geometry Clipping Detection in Video Game QA
Authors: Carlos Celemin , Benedict Wilkins , Adrián Barahona-Ríos , Saman Zadtootaghaj , Nabajeet Barman
View a PDF of the paper titled Evaluating VLMs for Autonomous Agent-Driven Geometry Clipping Detection in Video Game QA, by Carlos Celemin and 4 other authors
View PDF HTML (experimental)
Abstract: In this work, we study the use of Vision-Language Models (VLMs) for anomaly detection in an agent-driven game Quality Assurance (QA) pipeline focusing on geometry clipping. In this evaluation, a custom exploration agent navigates a game level to collect visual observations, while the automatic annotation pipeline provides frame-level clipping labels. This setup allows us to evaluate recent VLMs on a controlled anomaly detection task without manual annotation. We benchmark six recent VLMs (Gemini, GPT, Qwen, Gemma, Llama, and Ministral) under a zero-shot prompting setting and analyse their sensitivity to four prompt variants.
Our results show that while the VLMs can capture visual cues associated with geometry clipping, they all produce substantial false positives on visually ambiguous frames such as near-contact geometry and partial occlusions. Gemini-3.1-Flash achi
Extract from the original. Read the full piece at the source.
- agent
- language model
- prompt
- gpt
- gemini
The Agent Architect
One pattern, one tradeoff, one production failure story. A short weekly briefing for people building agentic systems.
Weekly email, one-click unsubscribe. We only use your address to send the briefing.