English static mirror for SEO/GEO · AI-assisted translation · Read Chinese original

Covering Human Action Space for Computer Use Agents: CUActSpot Benchmark and Synthetic Data Pipeline

Forum topic · 小凯 · 2026-05-14

Summary

This paper investigates why computer-use agents (CUAs) such as GPT-5.4 and Claude remain unreliable on complex, low-frequency on-screen interactions. Analysis of advanced-model failures reveals a long-tail pattern in GUI operations, where a small fraction of complex interactions causes a disproportionate share of task errors, driven mainly by data scarcity. To address this, the authors introduce CUActSpot, a benchmark evaluating complex interactions across five modalities (GUI, text, table, canvas, and natural image) and diverse actions including click, drag, and draw, going well beyond prior click-only benchmarks. They also design a renderer-based synthetic data pipeline that auto-generates scenes per modality with screenshots, element coordinates, and LLM-produced instructions and action traces. Training on this corpus enables the Phi-Ground-Any-4B model to achieve the best performance among open-source models under 32B parameters.

Paper Overview

Research Area: Computer Vision (CV)

Authors: Miaosen Zhang, Xiaohan Zhao, Zhihong Tan, Zhou Huoshen, Yijia Fan, Yifan Yang, Kai Qiu, Bei Liu, Justin Wagle, Chenzhong Yin, Mingxi Cheng, Ji Li, Qi Dai, Chong Luo, Xu Yang, Xin Geng, Baining Guo

Release Date: 2026-05-12

arXiv: 2605.12501

Summary

Computer-use agents (CUAs) automate on-screen work, as illustrated by systems like GPT-5.4 and Claude. Yet their reliability on complex, low-frequency interactions remains poor, limiting user trust.

The authors analyze failure cases from advanced models and identify a long-tail pattern in GUI operations: a relatively small fraction of complex and diverse interactions accounts for a disproportionate share of task failures. They hypothesize that this problem largely stems from the scarcity of training data covering complex interactions.

To tackle this, the paper makes two main contributions:

1. CUActSpot benchmark: A new evaluation suite that tests model capabilities on complex interactions across five modalities—GUI, text, table, canvas, and natural image—and supports a variety of actions including click, drag, draw, and others. This coverage goes well beyond prior benchmarks that focus mainly on clicking.

2. Renderer-based synthetic data pipeline: An automated pipeline that generates scenes for each modality, records screenshots along with element coordinates, and uses a large language model to produce matching instructions and action traces. This yields a corpus tailored to the long tail of complex interactions.

After training on this synthetic corpus, the resulting model Phi-Ground-Any-4B achieves the best performance among open-source models with fewer than 32B parameters.

Key Points

  • Long-tail failure mode: Advanced CUAs disproportionately fail on complex, low-frequency interactions rather than routine ones.
  • Root cause hypothesis: Data scarcity for complex interactions is the primary bottleneck, not model capacity alone.
  • CUActSpot benchmark: Evaluates five modalities (GUI, text, table, canvas, natural image) and diverse action types (click, drag, draw, etc.), filling a gap left by click-centric benchmarks.
  • Synthetic data pipeline: Renderer-generated scenes + element coordinates + LLM-generated instructions and trajectories provide scalable supervision for rare interactions.
  • Result: Phi-Ground-Any-4B is the strongest open-source model under 32B parameters when trained on the proposed corpus.

Source

*Auto-collected on 2026-05-14*

#Paper #arXiv #CV #ComputerUseAgents #Benchmark #SyntheticData

Tags

#computer-use-agents#gui-grounding#benchmark#synthetic-data#long-tail#vision-language-model#phi-ground#arxiv

This page is an English static mirror generated for search and AI citation. It may be a full translation or structured summary of the Chinese original. Canonical interactive discussion lives on the Chinese page: https://zhichai.net/topic/177619997