RunAgent SuperBrowser: A Theory of Autonomous Web Navigation Grounded in Human Browsing Behaviour

• URL: http://arxiv.org/abs/2606.09399v1
• Date: Mon, 08 Jun 2026 12:18:22 GMT
• Title: RunAgent SuperBrowser: A Theory of Autonomous Web Navigation Grounded in Human Browsing Behaviour
• Abstract summary: We present SUPERBROWSER, an autonomous web-navigation agent designed against a single guiding hypothesis.<n>A human reading a page does not retain every pixel they have seen; they look at a few candidate targets, decide on one, and remember only what is needed to keep the goal alive.<n>On the Mind2Web Hard benchmark, SUPERBROWSER attains 89.47% success, placing third overall and ahead of every published open/research browser-agent baseline by a large margin.
• Score: 0.0
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: We present SUPERBROWSER, an autonomous web-navigation agent designed against a single guiding hypothesis: a web agent should browse the way a person browses. A human reading a page does not retain every pixel they have seen; they look at a few candidate targets, decide on one, and remember only what is needed to keep the goal alive. We operationalize this perception-cognition-action triad as three coupled mechanisms. First, a vision-first bounding-box pipeline labels candidate interactive regions on every screenshot and feeds them, asynchronously prefetched, to the language model so that the "eye" precedes the "hand". Second, a three-role brain -- an Orchestrator that classifies and routes, a Planner that evaluates progress every few steps, and a Worker that emits per-step actions -- separates strategic from operational reasoning. Third, a structured Ledger stores only what a person would: the goal, the last three actions, a small set of facts and dead-ends, and a handful of checkpoints; a six-phase eviction loop systematically discards stale screenshots, state blobs, and reasoning traces from the live context. Action execution is a three-tier click cascade (Chrome DevTools Protocol to Puppeteer to scripted) with humanized Bezier motion, plus a chevron-aware bounding-box snapper that resolves the "small arrow beside a large label" ambiguity. On the Mind2Web Hard benchmark (66 tasks), SUPERBROWSER attains 89.47% success, placing third overall and ahead of every published open/research browser-agent baseline by a large margin. We argue that the gain comes not from any single trick but from the consistent application of a cognitive contract throughout the system.

Related papers

• String: An Agentic OS Where Every App Is a Markdown File [33.47077247999211]
  LLM agents have become a new class of software user, but every surface they work through was designed for someone else.<n>We present String, an open-source runtime that gives this user an interface of its own and treats the job as an operating-systems problem.
  arXiv  Detail & Related papers  (2026-08-28T07:44:41Z)
• BrowserForge: Scaling Web Episode via Parallel Browser Sandboxes [54.96895469970198]
  BrowserForge generates web interaction data at scale by driving many browser sandboxes in parallel over the open web.<n> BrowserForge couples three components: an open-web sourcing stage that exposes the agent to hundreds of thousands of real, openly reachable websites.<n>A rule-plus-model cleaning pipeline removes failed runs and rewrites the surviving reasoning into a single unified chain-of-thought style.
  arXiv  Detail & Related papers  (2026-08-25T17:35:42Z)
• Beyond Attack-Success Rate: Action-Graded Severity Scale for Tool-Using AI Agents [0.0]
  Agentic red-teaming benchmarks report whether an injected agent was compromised as a single bit: the attack succeeded, or it did not.<n>We argue that this binary attack-success rate discards the information a defender most needs, namely how harmful the resulting action was.<n>We introduce an action-graded harm rubric that scores an agent's tool-call trajectory on a seven-level ordinal scale (L0 to L6) according to whether the executed action was reversible, whether it crossed scope to reach another party, and whether it expanded privilege.
  arXiv  Detail & Related papers  (2026-07-08T14:38:01Z)
• Cookie-Bench: Continuous On-screen Key Interaction Evaluation for Web Generation [24.920344869492066]
  textbfdataname is an 11-domain, 54-leaf, 1,000-query WebDev benchmark spanning both static-presentation and interactive-application tasks.<n>textbfframename, grounded in Flavell's metacognitive monitoring, separates evidence accumulation from judgment across three stages.
  arXiv  Detail & Related papers  (2026-05-28T14:30:33Z)
• GTA: Generating Long-Horizon Tasks for Web Agents at Scale [82.43869456830664]
  We introduce a scalable framework, GTA, that integrates crawling, retrieval-based seeding, in-context generation, and automated quality control.<n>We instantiate the pipeline on over 50 websites covering e-commerce, government, forums, and news, with multilingual and multi-hop coverage.<n>Our contributions are three-fold: (i) formalizing multi-hop web-agent task generation, (ii) proposing an efficient and validated pipeline for automatic data creation, and (iii) releasing a dynamic benchmark with reproducible evaluation.
  arXiv  Detail & Related papers  (2026-05-28T01:05:50Z)
• Odysseys: Benchmarking Web Agents on Realistic Long Horizon Tasks [67.44219836008348]
  We introduce Odysseys: a benchmark of 200 long-horizon web tasks derived from real world browsing sessions evaluated on the live Internet.<n>We find that binary pass/fail evaluation is inadequate for long-horizon settings and introduce a rubric-based evaluation, annotating each Odysseys task with an average of 6.1 graded rubrics.<n>We find that the strongest models achieve a success rate of 44.5%, which leaves substantial room for future improvements.
  arXiv  Detail & Related papers  (2026-04-27T20:05:41Z)
• Visual Confused Deputy: Exploiting and Defending Perception Failures in Computer-Using Agents [9.457255218406333]
  Computer-using agents (CUAs) act directly on graphical user interfaces, yet their perception of the screen is often unreliable.<n>We formalize the visual confused deputy: a failure mode in which an agent authorizes an action based on a misperceived screen state.<n>To mitigate this threat, we propose the first guardrail that operates outside the agent's perceptual loop.
  arXiv  Detail & Related papers  (2026-03-16T01:31:32Z)
• WebOperator: Action-Aware Tree Search for Autonomous Agents in Web Environment [11.364916354844938]
  LLM-based agents operate in a greedy, step-by-step manner, selecting actions solely based on the current observation.<n>Without an explicit backtracking mechanism, agents struggle to correct errors or systematically explore alternative paths.<n>We introduce WebOperator, a tree-search framework that enables reliable backtracking and strategic exploration.
  arXiv  Detail & Related papers  (2025-12-14T13:56:54Z)
• BrowserAgent: Building Web Agents with Human-Inspired Web Browsing Actions [48.194688161526756]
  BrowserAgent operates directly on raw web pages via Playwright through a set of predefined browser actions.<n>We introduce an explicit memory mechanism to store key conclusions across steps, further enhancing the model's reasoning capabilities.
  arXiv  Detail & Related papers  (2025-10-12T15:43:37Z)
• BrowserArena: Evaluating LLM Agents on Real-World Web Navigation Tasks [51.803138848305814]
  We introduce BrowserArena, a live open-web agent evaluation platform that collects user-submitted tasks.<n>We identify three consistent failure modes: captcha resolution, pop-up banner removal, and direct navigation to URLs.<n>Our findings surface both the diversity and brittleness of current web agents.
  arXiv  Detail & Related papers  (2025-10-02T15:22:21Z)
• Auto-Intent: Automated Intent Discovery and Self-Exploration for Large Language Model Web Agents [68.22496852535937]
  We introduce Auto-Intent, a method to adapt a pre-trained large language model (LLM) as an agent for a target domain without direct fine-tuning. Our approach first discovers the underlying intents from target domain demonstrations unsupervisedly. We train our intent predictor to predict the next intent given the agent's past observations and actions.
  arXiv  Detail & Related papers  (2024-10-29T21:37:04Z)
• Self-Supervised Object Goal Navigation with In-Situ Finetuning [110.6053241629366]
  This work builds an agent that builds self-supervised models of the world via exploration. We identify a strong source of self-supervision that can train all components of an ObjectNav agent. We show that our agent can perform competitively in the real world and simulation.
  arXiv  Detail & Related papers  (2022-12-09T03:41:40Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.