SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2605.07134 · 2026

Region4Web: Rethinking Observation Space Granularity for Web Agents

Dongha Lee, Donguk Kwon, > Headphones

arXiv · PDF · Open in the Atlas

Code that ran

We lifted 5 functions out of this paper's own repositories and ran 4 of them in a sandbox. "Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results.

RepositoryRoleRan
ServiceNow/BrowserGym canonical 4 of 5
FunctionStatusWhere it lives
extract_data_items_from_aria Ran ServiceNow/BrowserGym/browsergym/core/src/browsergym/core/observation.py
pointer only (licence: NOASSERTION) · get_code("8194ee5c0231da57")
get_chatbox_classic Ran ServiceNow/BrowserGym/browsergym/core/src/browsergym/core/chat.py
pointer only (licence: NOASSERTION) · get_code("b938f1e73d583a18")
get_chatbox_modern Ran ServiceNow/BrowserGym/browsergym/core/src/browsergym/core/chat.py
pointer only (licence: NOASSERTION) · get_code("e01ad5d24888de64")
overlay_som Ran ServiceNow/BrowserGym/browsergym/core/src/browsergym/utils/obs.py
pointer only (licence: NOASSERTION) · get_code("b121272b1bd15e8b")
flatten_axtree_to_str Not yet run ServiceNow/BrowserGym/browsergym/core/src/browsergym/utils/obs.py
pointer only (licence: NOASSERTION) · get_code("567d81b57b0ac04f")

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Web agents perceive web pages through an observation space, yet its granularity has remained an underexamined design choice. Existing work treats observation at the same element-level granularity as the action space, leaving the page's functional organization implicit and forcing the agent to infer it from element-level signals at every step. We argue observation should instead operate at the granularity of functional regions, parts of the page that each serve a distinct purpose. We propose Region4Web, a framework that reorganizes the AXTree into functional regions through hierarchical decomposition and semantic abstraction, exposing the page's functional organization as the basis for page state understanding. Moreover, we propose PageDigest, a web-specific inference pipeline that delivers this regionlevel observation to the actor agent as a compact per-page digest that persists across steps. On the WebArena benchmark, PageDigest substantially reduces observation length while improving overall task success rate across diverse backbone large language models (LLMs) and established agent methods, regardless of backbone capacity. These results show that operating at the granularity of functional regions delivers a more compact and informative basis for the actor agent than element-level processing alone. Code is available at https://github.com/kwondu/region4web.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2605.07134")
get_code_for_paper("2605.07134")
have("2605.07134")

Connect an agent — have() is free.