用图搜索方法自动收集网页智能体数据，提升任务完成率。
Go-Browse: Training Web Agents with Structured Exploration
- 将数据收集转化为图搜索，跨任务复用信息提高效率
- 在WebArena上生成1万条成功轨迹，7B模型达21.7%成功率
- 适合想训练小型网页代理的研究者，效果超当前小模型
数字智能体普遍缺乏对环境的理解，例如网页浏览代理常在陌生网站中迷失，无法判断应访问哪些页面以达成目标。为此，我们提出Go-Browse，一种通过结构化探索大规模自动收集多样化、真实网页代理数据的方法。该方法将数据收集建模为图搜索，实现探索过程中的信息复用。我们在WebArena基准上实现了实例化，构建了一个包含10,000条成功任务解决轨迹和40,000次交互步骤的语料库，覆盖100个网址。在该数据集上微调一个70亿参数的语言模型，在WebArena基准上的成功率达到21.7%，优于GPT-4o mini 2.4%，且超过当前100亿以下参数模型的最先进水平2.9%。
原文摘要 · Abstract (English)
One of the fundamental problems in digital agents is their lack of understanding of their environment. For instance, a web browsing agent may get lost in unfamiliar websites, uncertain what pages must be visited to achieve its goals. To address this, we propose Go-Browse, a method for automatically collecting diverse and realistic web agent data at scale through structured exploration of web environments. Go-Browse achieves efficient exploration by framing data collection as a graph search, enabling reuse of information across exploration episodes. We instantiate our method on the WebArena benchmark, collecting a dataset of 10K successful task-solving trajectories and 40K interaction steps across 100 URLs. Fine-tuning a 7B parameter language model on this dataset achieves a success rate of 21.7% on the WebArena benchmark, beating GPT-4o mini by 2.4% and exceeding current state-of-the-art results for sub-10B parameter models by 2.9%.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。