[论文] [论文] Back to Basics: A Skill-Centric Approach to Generalist Web Agents

论文概要

研究领域: CV
作者: Minh Pham-Dinh, Md Farhan Ishmam, Hernan Yupanqui, Reaz Hassan Joarder, Kenneth Marino
发布时间: 2026-10-26
arXiv: 2610.35823

中文摘要

浏览器智能体缺的不只是更多练习，而是练对的东西。我们提出 MiniWeb——65 个相互连接网站、横跨 15 领域的拟真微缩网络，智能体像人一样导航、搜索与交易。369 个人工标注任务覆盖日常工作流，标注了归为 7 大技能族的 37 个可复用交互原语，能诊断「哪个技能失败」而非只问「任务是否成功」。对十个开源智能体的测试显示「拖拽与手势」最弱——无智能体解决超过 24% 的实例。为此提出 MiniWebGen：从 MiniWeb 自主生成经认证的、技能中心训练数据。仅用 3100 条认证轨迹微调 Qwen3.5-4B，在 13 个未见网站把单原语成功率提升 24.9 个百分点，每个技能族均有收益；但这些轨迹教的是短程交互，不能直接转化为完整任务成功。进而构建 MiniWebAgent：训练好的执行器与规划器配对，规划器把完整任务分解为技能级步骤。未见网站上任务成功率提升 19.4 个百分点，并迁移到外部基准：WebArena-Lite +9.3、Online-Mind2Web +16.1、WebVoyager +20.0，且未在它们的轨迹上训练。配方很简洁：直接练习可复用技能，再组合它们解决陌生任务。

原文摘要

Browser agents do not just need more practice. They need to practice the right things. We introduce MiniWeb, a realistic miniature of the web comprising 65 interconnected websites across 15 domains, where agents navigate, search, and transact as people do. Its 369 human-annotated tasks cover everyday workflows and label 37 reusable interaction primitives grouped into 7 skill families, allowing us to diagnose which skills fail rather than only whether a task succeeds. Benchmarking ten open-weight agents, we find that drag & gesture is consistently the weakest family, with no agent solving more than 24% of its instances. To address this gap, we introduce MiniWebGen, a framework that autonomously generates verified, skill-centric training data from MiniWeb. Fine-tuning Qwen3.5-4B on just 3.1K verified trajectories improves single-primitive success by 24.9pp across 13 unseen websites, with gains across every skill family. However, these trajectories teach short-horizon interactions, which do not directly translate to full-task success. We build on this result with MiniWebAgent, which pairs the trained executor with a planner that decomposes full tasks into the skill-level steps it learned to perform. This approach improves task success by 19.4pp points on unseen MiniWeb sites and transfers to external benchmarks, with gains of 9.3pp on WebArena-Lite, 16.1pp on Online-Mind2Web, and 20.0pp on WebVoyager, without training on their trajectories. We provide a simple recipe for compact browser agents: practice reusable skills directly, then compose them to solve unfamiliar tasks.

自动采集于 2026-10-28

#论文 #arXiv #CV #小凯