工作流

AppCopilot: Toward General, Accurate, Long-Horizon, and Efficient Mobile Agent

Jingru Fan, Yufan Dang 等 8 位作者

Describes a mobile-agent stack combining region-based coordinate correction, hierarchical task planning, multi-agent voting, and reuse of successful interaction traces.

AppAgent-Pro: A Proactive GUI Agent System for Multidomain Information Integration and User Assistance

Yuyang Zhao, Wentao Shi 等 4 位作者

Proposes a proactive GUI assistant that gathers information across domains in response to user requests and anticipated information needs.

BIMgent: Towards Autonomous Building Modeling via Computer-use Agents

Zihan Deng, Changyu Du 等 4 位作者

BIMgent plans software-specific workflows and executes GUI actions for building-information modeling. It evaluates multimodal building design and reconstruction tasks in professional authoring software.

macOSWorld: A Multilingual Interactive Benchmark for GUI Agents

Pei Yang, Hai Ci 等 3 位作者

macOSWorld provides 202 interactive tasks across 30 applications, with instructions and operating-system interfaces in five languages. It evaluates GUI agents on macOS workflows and includes a separate subset for deception-attack testing.

WorkForceAgent-R1: Incentivizing Reasoning Capability in LLM-based Web Agents via Reinforcement Learning

Yuchen Zhuang, Di Jin 等 6 位作者

Uses rule-based rewards for output structure and action correctness to train reasoning for enterprise web navigation.

ScienceBoard: Evaluating Multimodal Autonomous Agents in Realistic Scientific Workflows

Qiushi Sun, Zhoumianze Liu 等 21 位作者

ScienceBoard combines an interactive environment with professional scientific software and 169 human-curated tasks in areas including biochemistry, astronomy, and geoinformatics. It evaluates computer-use agents on visually rich research workflows and analyzes their execution limitations.

Towards Enterprise-Ready Computer Using Generalist Agent

Sami Marreed, Alon Oved 等 9 位作者

Describes the CUGA enterprise-agent development process, emphasizing iterative evaluation, failure analysis, and refinement on software-task benchmarks.

TheAgentCompany: Benchmarking LLM Agents on Consequential Real World Tasks

Frank F. Xu, Yufan Song 等 21 位作者

Simulates a software-company workplace to evaluate agents on professional tasks involving websites, programs, files, and coworker communication.

ScribeAgent: Towards Specialized Web Agents Using Production-Scale Workflow Data

Junhong Shen, Atishay Jain 等 7 位作者

Studies fine-tuning open language models on production web-workflow traces, including training choices for specialized web agents.

Agent Workflow Memory

Zora Zhiruo Wang, Jiayuan Mao 等 4 位作者

Agent Workflow Memory extracts reusable task routines and selectively supplies them to guide later web actions. It studies both offline workflow induction from training examples and online induction from test experience on Mind2Web and WebArena.

IDA: Breaking Barriers in No-code UI Automation Through Large Language Models and Human-Centric Design

Segev Shlomov, Avi Yaeli 等 7 位作者

IDA helps business users create web UI automation through guided demonstrations, a semantic programming model, and a teaching-oriented interface. A prototype user study examines automation creation with enterprise applications by users without programming backgrounds.

Spider2-V: How Far Are Multimodal Agents From Automating Data Science and Engineering Workflows?

Ruisheng Cao, Fangyu Lei 等 23 位作者

Benchmarks multimodal agents on data-science and engineering workflows involving enterprise applications, code, and GUI operations in real computer environments.

WorkArena++: Towards Compositional Planning and Reasoning-based Common Knowledge Work Tasks

Léo Boisvert, Megh Thakkar 等 9 位作者

Defines compositional enterprise workflows to evaluate planning and reasoning, with generation of ground-truth interaction traces.

WONDERBREAD: A Benchmark for Evaluating Multimodal Foundation Models on Business Process Management Tasks

Michael Wornow, Avanika Narayan 等 18 位作者

Introduces WONDERBREAD, a benchmark pairing human WebArena workflow demonstrations with screen recordings, action logs, and written procedures. Six evaluation tasks cover documentation, knowledge transfer, and process improvement beyond end-to-end execution.

SmartFlow: Robotic Process Automation using LLMs

Arushi Jain, Shubham Paliwal 等 5 位作者

Presents SmartFlow, which converts visible GUI elements into text for language-model action planning and executes the plan through a scripting engine. A collection of enterprise applications evaluates adaptation to different layouts and tasks.

Prompt2Task: Automating UI Tasks on Smartphones from Textual Prompts

Tian Huang, Chun Yu 等 7 位作者

Prompt2Task converts textual goals or procedures into smartphone UI automation using agents for intent interpretation, external-information handling, and execution. User feedback and accumulated task knowledge support subsequent task generation and operation.

WorkArena: How Capable Are Web Agents at Solving Common Knowledge Work Tasks?

Alexandre Drouin, Maxime Gasse 等 12 位作者

Introduces enterprise web tasks on ServiceNow and BrowserGym interfaces for evaluating browser-based agents.

PPTC-R benchmark: Towards Evaluating the Robustness of Large Language Models for PowerPoint Task Completion

Zekai Zhang, Yiduo Guo 等 5 位作者

Benchmarks robustness of API-based PowerPoint completion under sentence-level, semantic, and multilingual instruction perturbations and changes to the available API set.

MobileAgent: enhancing mobile control via human-machine interaction and SOP integration

Tinghe Ding

Combines operating-procedure context with user interaction to guide mobile task automation and handle personalized or sensitive requests.

InstructExcel: A Benchmark for Natural Language Instruction in Excel

Justin Payan, Swaroop Mishra 等 10 位作者

Introduces InstructExcel, a benchmark for generating Excel OfficeScripts from natural-language requests, built using scripts recorded from user actions and a collection of existing spreadsheets.

SteP: Stacked LLM Policies for Web Actions

Paloma Sodhi, S. R. K. Branavan 等 4 位作者

SteP dynamically composes specialized web-action policies through a stack that records the active chain of policy calls. It studies this control structure across WebArena, MiniWoB++, and CRM workflows.

SheetCopilot: Bringing Software Productivity to the Next Level through Large Language Models

Hongxin Li, Jingran Su 等 5 位作者

Introduces SheetCopilot, an agent that steers spreadsheet software via atomic virtual APIs, along with a 221-task spreadsheet manipulation benchmark.

Kite: Building Conversational Bots from Mobile Apps

Toby Jia-Jun Li, Oriana Riva

Derives task graphs from demonstrated Android app interactions and layers a conversational question-answer interface over those graphs. The evaluation examines task-model extraction and developer use of the resulting bot templates.

Reinforcement Learning on Web Interfaces Using Workflow-Guided Exploration

Evan Zheran Liu, Kelvin Guu 等 5 位作者

Uses workflows extracted from demonstrations to constrain reinforcement-learning exploration for interactive web tasks.

Koala: Capture, Share, Automate, Personalize Business Processes on the Web

Greg Little, Tessa A. Lau 等 6 位作者

Introduces Koala, a collaborative web automation system that records browser interactions as editable pseudo-natural-language scripts. Scripts are shared through a wiki and instantiated with personal or corporate data for reuse by other users.

ClawBench

Everyday-task benchmark on live websites that runs browser agents in containerized Chromium, blocks each final submission request for scoring by URL match and an LLM judge, and records five layers of run evidence.

OSWorld-Human

Human reference trajectories and scoring code for measuring the action efficiency of computer-use agents on OSWorld tasks.

SaaS-Bench

106 browser-driven business workflows across 23 self-hosted SaaS apps in six domains, scored by per-task verifiers that read each app's database and give partial credit.

Spider2-V

Visual agent evaluation on data-science and engineering workflows.

TheAgentCompany

Self-contained workplace environment for evaluating agents on professional tasks involving websites, files, code, and simulated coworkers.

WONDERBREAD

An offline benchmark of workflow documentation, quality assessment and knowledge transfer using human web-navigation demonstrations, action traces and written procedures.

WorkArena

Browser-based enterprise workflow tasks built around ServiceNow.

ScienceBoard evaluation trajectories

Archives of ScienceBoard agent evaluation runs, separated by agent configuration and screenshot, accessibility-tree or set-of-marks observation mode.

WorkDemo-350

Desktop demonstrations with task descriptions, videos, input logs, and action-linked screenshot evidence for studying reusable agent skills.

Agent Workflow Memory implementation

Research code that extracts reusable browser-task routines from demonstrations or prior executions, stores them as workflow memory and reuses them during subsequent navigation.

AskUI

Agentic test-automation platform running plain-language test cases from pixels rather than selectors on desktop apps, Citrix, in-car head units and mobile devices, deployable on-premise or air-gapped.

BrowserAct

Managed web scraping platform with reusable browser interaction workflows.

Bytebot

Open-source desktop agent that works inside its own Ubuntu Linux virtual desktop, with a web task interface, REST APIs for tasks and direct control, a takeover mode and password-manager logins.

ChatGPT Agent

Hosted agent combining browser interaction, tools, and computer-based tasks.

CHOP

Mobile automation framework that uses reusable basis subtasks to constrain and refine planning between user goals and executable actions.

Computer Use OOTB

Gradio desktop GUI agent for Windows and macOS that needs no Docker, using Claude 3.5 Computer Use over the API or ShowUI and UI-TARS run locally, with multi-display and phone remote control.

Cua

Monorepo behind Cua Spaces, which gives agents full macOS or Linux desktops, plus Cua Driver for operating native apps, Lume for Apple Silicon VMs, Cua Bench and the CUA-S1 decision models.

Manus

General agent platform with browser-based task execution.

MobileAgent (Alipay)

Mobile-control research implementation combining operating-procedure context with human interaction for personalized task execution.

MobileGPT

An Android client and Python-server agent that learns reusable app subtasks through exploration, selection, derivation and memory recall.

OpenAdapt

Tools for learning and automating workflows from desktop demonstrations.

Scribe

Workflow recording and documentation product adjacent to agent data collection.

SheetCopilot

Spreadsheet automation agent for Windows/Excel, with benchmark assets and a separately released headless Ubuntu evaluation backend.

ShowUI-Aloha

Desktop agent that records demonstrations, derives semantic action traces, and uses them to plan and execute related workflows.

Simular

Computer-use products and research from Simular.

Skyvern

Browser workflow automation that relies on vision LLMs and a swarm of agents instead of fixed XPath selectors, offered as a Playwright-compatible SDK and a no-code workflow builder.

Skyvern Cloud

Hosted Skyvern service for AI browser workflows such as invoice downloads and form filling, with built-in CAPTCHA and 2FA handling, proxy geotargeting, run explanations and JSON or CSV output.

Surfkit

Python toolkit and CLI for building, publishing and running multimodal agents that operate devices such as Ubuntu desktops, locally or on Kubernetes, with task trackers and an agent hub.

UI-TARS Desktop

Desktop app for a local GUI agent driven by UI-TARS and Seed-1.5-VL/1.6 models, offering local and remote operators for computers and browsers; the repository also ships the Agent TARS CLI and web UI.

Work2Skill

Collects human GUI demonstrations and inspects multimodal evidence to derive reusable agent workflow skills.

Workflow-Guided Exploration research code

Research code that learns browser-control policies by constraining exploration with workflows induced from demonstrations, with a Selenium-based MiniWoB environment and experiment harness.