已审阅并标注「网页导航」的研究与资产。
条目依据已审阅标签归类，同一主题不表示资源之间存在依赖关系。日期保留来源精度，注释保留原语言。
Web-CogReasoner separates factual and conceptual web knowledge from procedural reasoning, using a structured curriculum to train a multimodal agent. Web-CogDataset and Web-CogBench support study of these capabilities alongside web-navigation evaluation.
Magentic-UI combines browser interaction, code execution, file operations, and MCP tools in an interface for human-agent collaboration. It studies collaborative planning and task execution, action guards, and persistent memory through benchmark tasks, simulated users, user studies, and safety assessments.
Studies indirect prompt injection through HTML content consumed as accessibility trees. The evaluation tests how manipulated page observations redirect browser-agent actions on real websites.
R-VLM localizes GUI elements through zoomed-in region proposals and an intersection-over-union-aware training objective. It evaluates the grounding model on ScreenSpot and AgentStudio and studies its effect on navigation in AITW and Mind2Web.
SimpAgent prunes distracting interface elements with visual masks and compresses interaction history using consistency-based guidance. The study evaluates these two forms of context reduction during GUI navigation in web and mobile environments.
WebArxiv provides 510 static-snapshot tasks with deterministic answers for scholarly browsing, including constrained paper retrieval, content extraction, and cross-paper comparison. The study also tests dynamic retrieval of interaction history to address repetitive or incomplete agent reasoning.
LineRetriever uses a language model to select observation lines that matter for upcoming web-navigation actions. It reduces DOM or accessibility-tree context by considering the planning horizon instead of relying solely on semantic similarity.
Introduces environments that couple 3D physical-world simulations with functional web interfaces. The accompanying benchmark evaluates tasks requiring agents to coordinate online information and physical actions across cooking, navigation and shopping scenarios.
Mirage-1 organizes interaction trajectories into execution, core, and meta-skills, then uses them to guide Monte Carlo tree search during online GUI tasks. It introduces AndroidLH for long-horizon evaluation and tests the approach on Android and web-navigation benchmarks.
Introduces EconWebArena, a curated benchmark of economic-data tasks requiring live website navigation, multimodal interpretation, and precise extraction. It evaluates visual grounding, planning, and interaction design within multistep workflows.
Test-Time Interaction scales the interaction horizon so web agents can explore, backtrack and revise plans within a rollout. Its online reinforcement-learning curriculum adjusts rollout lengths and is evaluated on WebVoyager and WebArena.
Go-Browse frames web-environment exploration as graph search so that information can be reused across data-collection episodes. It collects 10,000 successful trajectories and 40,000 interaction steps on WebArena and uses them to fine-tune a web agent.
Describes Surfer-H, a web agent paired with the Holo1 vision-language model family trained on web content, synthetic examples, and agent-generated data. The work introduces WebClick for web-interface localization and evaluates browser task execution on WebVoyager.
DeepShop evaluates shopping agents with evolving query complexity across product attributes, filters and sorting preferences. Its evaluation combines checks of these individual requirements with overall task success.
Open CaptchaWorld evaluates multimodal agents on interactive browser-based CAPTCHA puzzles. Its benchmark distinguishes task success from the depth of visual reasoning and interaction needed to complete a puzzle.
Studies advertising delivery as an indirect-injection channel for web agents under a black-box threat model. AdInject evaluates how static ad content shaped by website context can redirect browser actions.
WebCoT turns reflection, look-ahead, branching and rollback procedures into chain-of-thought training rationales. Fine-tuning on these reconstructed trajectories is evaluated for web-agent self-improvement and transfer across navigation tasks.
Trains web agents through asynchronous multi-turn environment interaction, using task-success rewards to optimize complete trajectories.
Web-Shepherd is a process reward model for assessing individual steps of web-navigation trajectories. It introduces a collection of 40,000 step-level preference pairs with checklists and a separate benchmark for comparing reward models, then evaluates verification during WebArena-lite navigation.
Outlines an end-to-end threat model and empirical security evaluation across architectural layers of autonomous browser agents.
RegionFocus adds visual test-time scaling by zooming into selected screenshot regions and maintaining an image-based map of landmarks and actions. It evaluates this grounding procedure with UI-TARS and Qwen2.5-VL on ScreenSpot-Pro and WebVoyager.
Combines iterative web search and link-following with interleaved reasoning and report drafting for deep research. Its reported web explorer retrieves page content through a crawler rather than controlling a visual browser interface.
WebEvolver trains a co-evolving world model to predict web observations and generate additional agent-training trajectories. The same model supports look-ahead action selection at inference, with evaluation on live web-task benchmarks.
Introduces WASP, an end-to-end benchmark for prompt-injection security in web agents. Its evaluation separates partial attack progress from completion of the attacker’s full objective.
WebRollback gives a web agent an explicit action for returning to an earlier state in its navigation trajectory. It studies this recovery mechanism on two live web-navigation benchmarks under both zero-shot and fine-tuned settings.
AgentRewardBench evaluates automated web-agent judges against expert reviews of 1,302 trajectories from five benchmarks. The annotations cover completion, side effects, and repetition, enabling comparison of LLM judges and rule-based task evaluators.
SkillWeaver explores websites, practices interactions, and converts the resulting experience into a reusable library of callable skill APIs. Its WebArena and live-website experiments examine skill refinement, composition, and transfer from stronger agents to weaker ones.
Adapts web agents to execute natural-language test steps, verify assertions, and issue test verdicts. The study introduces 113 passing and failing test cases across three offline web applications and compares two agent implementations with qualitative failure analysis.
Introduces Online-Mind2Web, with 300 tasks across 136 live websites, and examines how benchmark design affects assessments of web agents. It also develops an LLM judge for task completion and compares its judgments with human evaluation.
Surveys foundation-model web agents through their architectures, training methods and trustworthiness. It organizes research on completing browser tasks from user instructions and identifies directions for further development.
BEARCUBS provides 111 live-web information-seeking questions with short answers and human-validated browsing trajectories. Its tasks require multimodal computer interaction, including video understanding and navigation of interactive 3D content.
Introduces a modular visual web-agent stack with planning and memory components, exposed through a remote-browser application and Chrome extension.
Analyzes web-agent vulnerabilities at the component level, examining goal placement in prompts, multistep action generation, and observations as sources of increased susceptibility.
Synthesizes multimodal web trajectories through exploration and refinement, then trains a browser agent to study the effect of scaling interaction data.
Couples large- and small-model web-agent exploration with trajectory synthesis and distillation to improve both model types iteratively.
Introduces CowPilot, a web-navigation framework where agents suggest actions and users can pause, reject, override, or resume them. It evaluates autonomous and collaborative use through task success and human effort.
Introduces WebWalkerQA and an exploration-and-critique framework for gathering information by traversing multiple levels of a website.
Combines autonomous task proposal, web interaction, and visual success evaluation to discover skills and improve browser-agent policies.
Uses web-element preference optimization with sampled negative HTML elements to align high-level user intent with navigation actions.
Converts public tutorials into task specifications and executes them with an agent to synthesize multimodal interaction trajectories.
Studies fine-tuning open language models on production web-workflow traces, including training choices for specialized web agents.
Adapts multimodal web agents to unfamiliar websites using a small number of human demonstrations through prompting or meta-learning.
Surveys virtual agents for browsers, applications and operating systems, organizing prior work by task type, observation and action spaces, model strategies and evaluation methods. It examines screen-based interaction and keyboard, mouse and touch control, alongside transferability and long-sequence decision-making challenges.
Trains web agents with an evolving task curriculum, outcome supervision, and reinforcement-learning updates intended to manage policy drift.
Analyzes how interaction history and webpage representation affect multiturn web-agent generalization across unfamiliar websites, task categories, and geographic settings.
Discovers compact intents from demonstrations and uses intent predictions to guide a language-model web agent's decisions and exploration.
Develops multimodal browser agents through repeated cycles of real-web exploration, trajectory feedback, and policy optimization.
Compares browsing-only, API-only, and hybrid web agents on WebArena, examining how access to application APIs changes task execution alongside browser interaction.
Studies how simplifying and aligning a web agent's observations and actions with language-model capabilities affects task completion.
Generates browser-agent demonstrations through autonomous exploration and retrospective task labeling, pruning trajectories that lack meaningful subtasks.
Studies text and visual validation for web agents and uses the resulting validator to detect failures and guide workflow self-refinement.
Synatra converts indirect procedural knowledge, including human-oriented tutorials, into direct demonstrations for digital agents. It trains a 7B CodeLlama on 100,000 synthesized demonstrations and evaluates the resulting policy on Mind2Web, MiniWoB++, and WebArena.
Introduces Steward, a natural-language browser automation system that repeatedly plans and executes website actions. The paper examines completion detection, state representation, responsiveness, and caching in real-site tasks.
Evaluates privacy-directed webpage instruction injection against generalist web agents using PII-related interaction steps from Mind2Web, including attempts to expose specific user information or entire requests.
Frames web-application exploration as functionality-oriented question answering, generating navigation actions from multimodal context.
Agent Workflow Memory extracts reusable task routines and selectively supplies them to guide later web actions. It studies both offline workflow induction from training examples and online induction from test experience on Mind2Web and WebArena.
WebPilot separates global subtask planning from local Monte Carlo tree search during web execution. It continually refines the high-level plan and adapts local search to uncertain transitions and incomplete observations, evaluated on WebArena and MiniWoB++.
Introduces realistic open-web tasks and the SeePlanAct agent, analyzing the limitations of browsing and retrieval in time-consuming assistant work.
Agent-E combines hierarchical web-task control with DOM distillation and observations of interface changes. WebVoyager experiments inform a broader discussion of primitive skills, observation reduction, and self-improvement in agent architecture.
Applies best-first search to agent actions within interactive web environments, allowing exploration and multistep planning at inference time. WebArena and VisualWebArena experiments examine search behavior and test-time computation.
第 121–180 条，共 306 条资源。
通过主题导出清单获取完整记录和证据。克隆目录后，可将主题记录直接交给标准 RAG 与引用导出工具。
