LineRetriever: Planning-Aware Observation Reduction for Web Agents

While large language models have demonstrated impressive capabilities in web navigation tasks, the extensive context of web pages, often represented as DOM or Accessibility Tree (AxTree) structures, frequently exceeds model context limits. Current approaches like bottom-up truncation or embedding-based retrieval lose critical information about page state and action history. This is particularly problematic for adaptive planning in web agents, where understanding the current state is essential for determining future actions. We hypothesize that embedding models lack sufficient capacity to capture plan-relevant information, especially when retrieving content that supports future action prediction. This raises a fundamental question: how can retrieval methods be optimized for adaptive planning in web navigation tasks? In response, we introduce textitLineRetriever, a novel approach that leverages a language model to identify and retrieve observation lines most relevant to future navigation steps. Unlike traditional retrieval methods that focus solely on semantic similarity, textitLineRetriever explicitly considers the planning horizon, prioritizing elements that contribute to action prediction. Our experiments demonstrate that textitLineRetriever can reduce the size of the observation at each step for the web agent while maintaining consistent performance within the context limitations.