通过渐进式训练课程，提升网页导航模型的鲁棒性与泛化能力。
From Imitation to Discrimination: Progressive Curriculum Learning for Robust Web Navigation
- 构建含59万样本的Triton数据集，融合结构语义难例挖掘与双代理验证
- 三阶段模型在Mind2Web上达58.7%步成功率，超越GPT-4.5和Claude-4.5超16%
- 适合需要高鲁棒性网页自动化任务的研究者与开发者
文本驱动的网页代理虽具计算效率优势，但真实网页中HTML的噪声与异构性使其难以构建鲁棒代理。标准监督微调（SFT）在排斥密集页面中看似合理但错误的元素、以及泛化至未见网站布局方面表现不足。为此，本文提出Triton数据集（59万实例），通过结构-语义难例挖掘显式生成拓扑相似干扰项，并采用双代理共识管道合成多样化跨域任务并严格验证。在此基础上，设计渐进式训练课程，产出三个模型：Triton-SFT-32B用于基础模仿，Triton-ORPO-32B通过奇比偏好优化实现鲁棒判别，Triton-GRPO-32B则通过组相对策略优化保障长序列一致性。在Mind2Web上的实证表明，Triton-GRPO-32B作为开源模型达到58.7%的步成功率，显著优于GPT-4.5（42.4%）与Claude-4.5（41.4%），验证了专用数据课程优于单纯参数规模对网页导航的重要性。
原文摘要 · Abstract (English)
Text-based web agents offer computational efficiency for autonomous web navigation, yet developing robust agents remains challenging due to the noisy and heterogeneous nature of real-world HTML. Standard Supervised Fine-Tuning (SFT) approaches fail in two critical dimensions: they lack discrimination capabilities to reject plausible but incorrect elements in densely populated pages, and exhibit limited generalization to unseen website layouts. To address these challenges, we introduce the Triton dataset (590k instances) and a progressive training curriculum. Triton is constructed via Structural-Semantic Hard Negative Mining, which explicitly mines topologically similar distractors, and a Dual-Agent Consensus pipeline that synthesizes diverse cross-domain tasks with strict verification. Building upon this foundation, our progressive curriculum produces three models: Triton-SFT-32B for basic imitation, Triton-ORPO-32B for robust discrimination via Odds Ratio Preference Optimization, and Triton-GRPO-32B for long-horizon consistency through Group Relative Policy Optimization. Empirical evaluation on Mind2Web demonstrates that Triton-GRPO-32B achieves state-of-the-art performance among open-source models with 58.7% Step Success Rate, surpassing GPT-4.5 (42.4%) and Claude-4.5 (41.4%) by over 16%, validating that specialized data curriculum outweighs raw parameter scale for web navigation.
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。