Programming with Pixels: Can Computer-Use Agents do Software Engineering?

• URL: http://arxiv.org/abs/2502.18525v2
• Date: Fri, 03 Oct 2025 02:09:59 GMT
• Title: Programming with Pixels: Can Computer-Use Agents do Software Engineering?
• Authors: Pranjal Aggarwal, Sean Welleck,
• Abstract summary: $textttProgramming with Pixels$ (PwP) is the first comprehensive computer-use environment for software engineering.<n>PwP establishes software engineering as a natural domain for benchmarking whether generalist computer-use agents can reach specialist-level performance.
• Score: 24.011063667060792
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Computer-use agents (CUAs) hold the promise of performing a wide variety of general tasks, but current evaluations have primarily focused on simple scenarios. It therefore remains unclear whether such generalist agents can automate more sophisticated and specialized work such as software engineering (SWE). To investigate this, we introduce $\texttt{Programming with Pixels}$ (PwP), the first comprehensive computer-use environment for software engineering, where agents visually control an IDE to perform diverse software engineering tasks. To enable holistic evaluation, we also introduce \texttt{PwP-Bench}, a benchmark of 15 existing and new software-engineering tasks spanning multiple modalities, programming languages, and skillsets. We perform an extensive evaluation of state-of-the-art open-weight and closed-weight CUAs and find that when interacting purely visually, they perform significantly worse than specialized coding agents. However, when the same CUAs are given direct access to just two APIs-file editing and bash operations-performance jumps, often reaching the levels of specialized agents despite having a task-agnostic design. Furthermore, when given access to additional IDE tools via text APIs, all models show further gains. Our analysis shows that current CUAs fall short mainly due to limited visual grounding and the inability to take full advantage of the rich environment, leaving clear room for future improvements.PwP establishes software engineering as a natural domain for benchmarking whether generalist computer-use agents can reach specialist-level performance on sophisticated tasks. Code and data released at https://programmingwithpixels.com

Related papers

• Computer-Using World Model [58.59112582915026]
  We introduce the Computer-Using World Model (CUWM), a world model for desktop software that predicts the next user interface (UI) state.<n> CUWM first predicts a textual description of agent-relevant state changes, and then realizes these changes visually to synthesize the next screenshot.<n>We evaluate CUWM via test-time action search, where a frozen agent uses the world model to simulate and compare candidate actions before execution.
  arXiv  Detail & Related papers  (2026-02-19T13:48:29Z)
• IDE-Bench: Evaluating Large Language Models as IDE Agents on Real-World Software Engineering Tasks [0.37823923040445995]
  We present a Dockerized test harness that goes beyond raw terminal execution.<n>We provide high-level abstractions for search, structured file editing, and tools for testing full-stack applications.
  arXiv  Detail & Related papers  (2026-01-28T02:06:37Z)
• ABC-Bench: Benchmarking Agentic Backend Coding in Real-World Development [72.4729759618632]
  We introduce ABC-Bench, a benchmark to evaluate agentic backend coding within a realistic, executable workflow.<n>We curated 224 practical tasks spanning 8 languages and 19 frameworks from open-source repositories.<n>Our evaluation reveals that even state-of-the-art models struggle to deliver reliable performance on these holistic tasks.
  arXiv  Detail & Related papers  (2026-01-16T08:23:52Z)
• Computer-Use Agents as Judges for Generative User Interface [142.75272102498806]
  ComputerUse Agents (CUA) are becoming increasingly capable of autonomously operating digital environments through Graphical User Interfaces (GUI)<n>Most GUI remain designed primarily for humans to adopt human-oriented behaviors that are unnecessary for efficient task execution.<n>This raises a fundamental question: Can CUA as judges to assist Coder for automatic GUI design?
  arXiv  Detail & Related papers  (2025-11-19T16:00:02Z)
• Unified Software Engineering agent as AI Software Engineer [14.733475669942276]
  Large Language Model (LLM) technology has raised expectations for automated coding.<n>In this paper, we seek to understand this question by developing a Unified Software Engineering agent or USEagent.<n>We envision USEagent as the first draft of a future AI Software Engineer which can be a team member in future software development teams involving both AI and humans.
  arXiv  Detail & Related papers  (2025-06-17T16:19:13Z)
• Coding Agents with Multimodal Browsing are Generalist Problem Solvers [48.938445118630284]
  OpenHands-Versa is a generalist AI agent built with a modest number of general tools.<n>We show how existing state-of-the-art multi-agent systems fail to generalize beyond their target domains.
  arXiv  Detail & Related papers  (2025-06-03T15:50:55Z)
• ThinkGeo: Evaluating Tool-Augmented Agents for Remote Sensing Tasks [54.52092001110694]
  ThinkGeo is a benchmark designed to evaluate tool-augmented agents on remote sensing tasks via structured tool use and multi-step planning.<n>Inspired by tool-interaction paradigms, ThinkGeo includes human-curated queries spanning a wide range of real-world applications.<n>Our analysis reveals notable disparities in tool accuracy and planning consistency across models.
  arXiv  Detail & Related papers  (2025-05-29T17:59:38Z)
• InfantAgent-Next: A Multimodal Generalist Agent for Automated Computer Interaction [35.285466934451904]
  This paper introduces textscInfantAgent-Next, a generalist agent capable of interacting with computers in a multimodal manner.<n>Unlike existing approaches that either build intricate around a single large model or only provide modularity, our agent integrates tool-based and pure vision agents.
  arXiv  Detail & Related papers  (2025-05-16T05:43:27Z)
• LLM Agents Making Agent Tools [2.5529148902034637]
  Tool use has turned large language models (LLMs) into powerful agents that can perform complex multi-step tasks.<n>We propose ToolMaker, a novel agentic framework that autonomously transforms papers with code into LLM-compatible tools.<n>Given a short task description and a repository URL, ToolMaker autonomously installs required dependencies and generates code to perform the task.
  arXiv  Detail & Related papers  (2025-02-17T11:44:11Z)
• RepoGraph: Enhancing AI Software Engineering with Repository-level Code Graph [63.87660059104077]
  We present RepoGraph, a plug-in module that manages a repository-level structure for modern AI software engineering solutions.<n>RepoGraph substantially boosts the performance of all systems, leading to a new state-of-the-art among open-source frameworks.
  arXiv  Detail & Related papers  (2024-10-03T05:45:26Z)
• MetaTool: Facilitating Large Language Models to Master Tools with Meta-task Augmentation [25.360660222418183]
  We present MetaTool, a novel tool learning methodology designed to generalize across any reusable toolset. By incorporating meta-task data into task-oriented training, our method significantly enhances the performance of open-source Large Language Models.
  arXiv  Detail & Related papers  (2024-07-15T10:15:41Z)
• Agentless: Demystifying LLM-based Software Engineering Agents [12.19683999553113]
  We build Agentless -- an agentless approach to automatically solve software development problems. Compared to the verbose and complex setup of agent-based approaches, Agentless employs a simplistic three-phase process of localization, repair, and patch validation. Our results on the popular SWE-bench Lite benchmark show that surprisingly the simplistic Agentless is able to achieve both the highest performance and low cost.
  arXiv  Detail & Related papers  (2024-07-01T17:24:45Z)
• CAAP: Context-Aware Action Planning Prompting to Solve Computer Tasks with Front-End UI Only [21.054681757006385]
  We propose an agent that perceives its environment solely through screenshot images.<n>By leveraging the reasoning capability of the Large Language Models, we eliminate the need for large-scale human demonstration data.<n>Agent achieves an average success rate of 94.5% on MiniWoB++ and an average task score of 62.3 on WebShop.
  arXiv  Detail & Related papers  (2024-06-11T05:21:20Z)
• Morescient GAI for Software Engineering (Extended Version) [2.4861619769660637]
  We present a vision and roadmap for how "Morescient" GAI models can be engineered, evolved and disseminated according to the principles of open science.
  arXiv  Detail & Related papers  (2024-06-07T07:38:33Z)
• SWE-agent: Agent-Computer Interfaces Enable Automated Software Engineering [79.07755560048388]
  SWE-agent is a system that facilitates LM agents to autonomously use computers to solve software engineering tasks. SWE-agent's custom agent-computer interface (ACI) significantly enhances an agent's ability to create and edit code files, navigate entire repositories, and execute tests and other programs. We evaluate SWE-agent on SWE-bench and HumanEvalFix, achieving state-of-the-art performance on both with a pass@1 rate of 12.5% and 87.7%, respectively.
  arXiv  Detail & Related papers  (2024-05-06T17:41:33Z)
• OSWorld: Benchmarking Multimodal Agents for Open-Ended Tasks in Real Computer Environments [87.41051677852231]
  We introduce OSWorld, the first-of-its-kind scalable, real computer environment for multimodal agents. OSWorld can serve as a unified, integrated computer environment for assessing open-ended computer tasks. We create a benchmark of 369 computer tasks involving real web and desktop apps in open domains, OS file I/O, and spanning multiple applications.
  arXiv  Detail & Related papers  (2024-04-11T17:56:05Z)
• AgentStudio: A Toolkit for Building General Virtual Agents [57.02375267926862]
  General virtual agents need to handle multimodal observations, master complex action spaces, and self-improve in dynamic, open-domain environments.<n>AgentStudio provides a lightweight, interactive environment with highly generic observation and action spaces.<n>It integrates tools for creating online benchmark tasks, annotating GUI elements, and labeling actions in videos.<n>Based on our environment and tools, we curate an online task suite that benchmarks both GUI interactions and function calling with efficient auto-evaluation.
  arXiv  Detail & Related papers  (2024-03-26T17:54:15Z)
• What Are Tools Anyway? A Survey from the Language Model Perspective [67.18843218893416]
  Language models (LMs) are powerful yet mostly for text generation tasks. We provide a unified definition of tools as external programs used by LMs. We empirically study the efficiency of various tooling methods.
  arXiv  Detail & Related papers  (2024-03-18T17:20:07Z)
• ScreenAgent: A Vision Language Model-driven Computer Control Agent [17.11085071288194]
  We build an environment for a Vision Language Model (VLM) agent to interact with a real computer screen. Within this environment, the agent can observe screenshots and manipulate the Graphics User Interface (GUI) by outputting mouse and keyboard actions. We construct the ScreenAgent dataset, which collects screenshots and action sequences when completing a variety of daily computer tasks.
  arXiv  Detail & Related papers  (2024-02-09T02:33:45Z)
• Planning, Creation, Usage: Benchmarking LLMs for Comprehensive Tool Utilization in Real-World Complex Scenarios [93.68764280953624]
  UltraTool is a novel benchmark designed to improve and evaluate Large Language Models' ability in tool utilization. It emphasizes real-world complexities, demanding accurate, multi-step planning for effective problem-solving. A key feature of UltraTool is its independent evaluation of planning with natural language, which happens before tool usage.
  arXiv  Detail & Related papers  (2024-01-30T16:52:56Z)
• ControlLLM: Augment Language Models with Tools by Searching on Graphs [97.62758830255002]
  We present ControlLLM, a novel framework that enables large language models (LLMs) to utilize multi-modal tools for solving real-world tasks. Our framework comprises three key components: (1) a textittask decomposer that breaks down a complex task into clear subtasks with well-defined inputs and outputs; (2) a textitThoughts-on-Graph (ToG) paradigm that searches the optimal solution path on a pre-built tool graph; and (3) an textitexecution engine with a rich toolbox that interprets the solution path and runs the
  arXiv  Detail & Related papers  (2023-10-26T21:57:21Z)
• WebArena: A Realistic Web Environment for Building Autonomous Agents [92.3291458543633]
  We build an environment for language-guided agents that is highly realistic and reproducible. We focus on agents that perform tasks on the web, and create an environment with fully functional websites from four common domains. We release a set of benchmark tasks focusing on evaluating the functional correctness of task completions.
  arXiv  Detail & Related papers  (2023-07-25T22:59:32Z)
• ART: Automatic multi-step reasoning and tool-use for large language models [105.57550426609396]
  Large language models (LLMs) can perform complex reasoning in few- and zero-shot settings. Each reasoning step can rely on external tools to support computation beyond the core LLM capabilities. We introduce Automatic Reasoning and Tool-use (ART), a framework that uses frozen LLMs to automatically generate intermediate reasoning steps as a program.
  arXiv  Detail & Related papers  (2023-03-16T01:04:45Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.