Read More, Think More: Revisiting Observation Reduction for Web Agents
This work addresses the problem of efficient web agent design for AI researchers and practitioners, offering practical guidelines for adaptive observation representation, though it is incremental in refining existing methods.
The study revisits observation reduction for web agents, finding that optimal web page representations depend on model capability and thinking token budget: compact observations work better for lower-capability models, while detailed HTML benefits higher-capability models, with performance gains up to 15% in some settings when using HTML with more thinking tokens.
Web agents based on large language models (LLMs) rely on observations of web pages -- commonly represented as HTML -- as the basis for identifying available actions and planning subsequent steps. Prior work has treated the verbosity of HTML as an obstacle to performance and adopted observation reduction as a standard practice. We revisit this trend and demonstrate that the optimal observation representation depends on model capability and thinking token budget: (1) compact observations (accessibility trees) are preferable for lower-capability models, while detailed observations (HTML) are advantageous for higher-capability models; moreover, increasing thinking tokens further amplifies the benefit of HTML. (2) Our error analysis suggests that higher-capability models exploit layout information in HTML for better action grounding, while lower-capability models suffer from increased hallucination under longer inputs. We also find that incorporating observation history improves performance across most models and settings, and a diff-based representation offers a token-efficient alternative. Based on these findings, we suggest practical guidelines: adaptively select observation representations based on model capability and thinking token budget, and incorporate observation history using diff-based representations.