不同观察模式各有优劣，但动态切换反而不如固定一个好模式。
Routing Is Least Learnable Where It Is Most Valuable: Bounds on Representation Routing for Web Agents
- 测试六种观察模式在八组场景下的表现，发现每种模式都有独特短板。
- 重跑相同模式会改变12%-14%结果，新模式增益有限。
- 路由策略难胜固定模式，因弱代理缺乏足够标注来训练路由
Web代理通过文本、像素或两者结合观察浏览器，通常固定使用一种观察方式。我们在VisualWebArena和WebArena上评估六种观察模式在八组站点-模型组合中的表现，探究按任务选择观察模式的收益。结果表明：各模式互补，各自解决其他模式无法处理的任务，失败模式也不同，最佳选择随任务集变化。看似理想的‘最优模式选择’虽有吸引力，但被运行间噪声夸大——同一模式重复运行可导致12%-14%的结果波动，因此新增一个模式的收益几乎等同于重新运行已有模式。真正可持续的收益是：仅将所有模式均无法解决的任务交给最便宜的模式，可在不降低成功率的前提下降低9.5%-30.6%成本（8/8场景）。我们测试了五种路由策略（模式选择、强模式投入时机、零成本规则、置信度级联、成本分层），但无一能稳定超越固定一个优选模式的方案；唯一例外出现在数据最稀疏的场景中。核心障碍在于：路由监督依赖于代理的成功率——代理越弱，获得的标签越少，而此时正是路由最需要的地方。该限制属于当前代理能力而非路由本身。标签供给与路由机会高度相关（跨场景相关系数达0.95），更强的代理可打破此瓶颈，本文还报告了重跑噪声区间及完整测量流程。
原文摘要 · Abstract (English)
Web agents observe a browser through text, pixels, or both, and the choice is usually fixed once for all tasks. We measure six observation modes across eight site-model combinations (cells) on VisualWebArena and WebArena and ask what choosing per task would buy. The modes are complementary: each solves tasks the others miss, they fail in structurally different ways, and the best choice reverses between task sets. The obvious prize, an oracle that picks a winning mode for every task, looks large but is inflated by run-to-run noise: rerunning the same mode on the same tasks changes 12-14% of outcomes, so a second run of a mode already in hand gains about as much as adding a new one. What survives is a cost bound: sending only the tasks no mode solves to the cheapest mode cuts cost by 9.5-30.6% in 8 of 8 cells at unchanged success. We then test five routing policies (picking the mode, deciding when to spend on the strong mode, a zero-cost rule read off the task text, a confidence cascade, and pooled cost tiers), and none robustly beats simply fixing one well-chosen mode; the one exception is a fragile result in our sparsest cell. The central obstruction is that routing supervision is produced at the agent's success rate: the weaker the agent, the fewer labels a router gets, exactly where routing would be most valuable. This limit belongs to today's agents rather than to routing itself. Label supply and routing opportunity rise together (correlation 0.95 across cells), so a stronger agent can overturn the result, and we report the rerun noise bands and the full measurement protocol.
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。