Focused Omni-Modal GUI Agents
FocusGUI learns step-specific weights over screenshot, audio, video and interaction-context channels, injecting a compact residual into a frozen decoder without pruning its inputs.
Android
界面理解
已审阅并标注「界面理解」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。
STP-GUI: Spatially-Aware Token Pruning for Efficient Web-GUI Agents
Introduces STP-GUI, which selects spatially coherent screenshot windows conditioned on the instruction to reduce visual tokens while retaining useful interface structure.
网页Android
Towards Safe Autonomy for Computer Use
Presents SA-CUA, separating trusted and untrusted visual content using accessibility-derived masking rules and a controlled channel for verifying information before endorsement.
MHVF: Training-Free GUI Grounding via Verification-Guided Multi-Hypothesis Search
Introduces MHVF, a training-free grounding method that explores several candidate screen regions, verifies them in local views, and uses those checks to guide further search.
网页
Seg2Click: Learning Pixel-Level Action Masks for GUI Visual Grounding
Recasts GUI grounding as predicting a mask of valid click locations from bounding-box supervision, then derives a click point with a deterministic selection rule.
CaptchaAgent: Verifiable Visual Reasoning for CAPTCHA Solving
Presents CaptchaAgent, which builds an explicit visual representation of objects and relations, refines ambiguous regions, and checks counterfactual consistency before acting on interactive CAPTCHAs.
网页
When Charts Deceive Web Agents: A Paired Benchmark for Misleading Visualization-Grounded Tasks
Introduces MisVisAgentBench, a paired executable benchmark that changes chart evidence while holding web tasks fixed, to trace how misleading visualizations affect agent decisions and downstream actions.
网页
Recurrent Latent Visual Search for GUI Grounding
Introduces ReLaViS, which performs recurrent visual search inside a grounding model by repeatedly querying screenshot tokens and feeding the resulting spatial evidence back into the model. Coarse-to-fine supervision is constructed from existing interface-element annotations.
OpenVAM: Open-World Visual Attention Modeling with VLMs
Combines dense saliency prediction with explanations of attended elements, including webpage layouts; evaluates human-attention agreement and textual rationales rather than agent actions.
网页相关研究
Invisible in Space, Visible in Time: Motion Vision CAPTCHA against GUI Agents
MVCAP-Bench tests live browser CAPTCHAs whose foreground meaning emerges through motion against dynamic backgrounds. A foreground-only control separates temporal camouflage difficulty from answer format and browser interaction.
网页
RankGround: Efficient High-Resolution GUI Grounding via Lightweight Reranker-Guided Crop Selection
RankGround trains a lightweight reranker to select a promising screenshot crop before one VLM localization call. Ranking supervision uses containment and boundary-aware augmentation, with pointwise and listwise training stages.
When Agents See Differently: Exposing UI Desynchronization Threats in Mobile Agents
Investigates mismatches between what humans see on a mobile display and what agents receive through screenshots or accessibility metadata. An automated evaluation framework embeds instruction-agnostic perturbations into functional APK clones and compares static and dynamic agent deviations with human noticeability.
Android
PriMobiBench: Characterizing Visual Privacy Leakage in VLM-Driven Mobile GUI Agents
PriMobiBench evaluates direct privacy leakage and user profiling from mobile GUI screenshot trajectories, supported by the MobiLeak trace dataset. It also tests masking sensitive, task-irrelevant interface elements before cloud processing and measures the resulting privacy and utility trade-off.
LLaDA-UI: Bringing Block-wise Diffusion to Vision-Language GUI Agents
LLaDA-UI adapts block-wise diffusion generation to a vision-language GUI agent through multimodal pretraining and GUI supervised fine-tuning. It evaluates grounding and navigation across multiple platform settings.
网页桌面
TRACE: Trajectory-robust Admission with Evidence Ordering for Efficient GUI Agents
TRACE orders screenshot evidence for monotonic visual-token reduction using layout priors, instruction relevance and feature novelty. Spatial token reserves and incremental KV contraction aim to preserve useful evidence for future GUI targets without repeatedly encoding retired frames.
Reflection with Action-Induced Visual Differences for Desktop GUI Agents
Introduces Evidence-First Reflection, which separates extraction of action-induced screen changes from judgment of whether an operation achieved its objective. It filters visual differences into relevant evidence before a desktop agent’s reflector decides how execution should continue.
桌面Windows
GUI-Primitives: Diagnosing Spatial Reasoning Failures in Vision-Language GUI Grounding
Introduces GUI-Primitives, a paired-instruction benchmark that changes a spatial relation while holding the screenshot and anchor fixed. Its diagnostics separate failure to localize either candidate from failure to choose the candidate satisfying the relation.
Lexical Coupling in GUI Element Grounding: Sentence Embeddings Track Labels across Mobile and Web
Studies whether text-embedding GUI grounding scores reflect semantic matching or recovery of visible labels. Same-screen ranking comparisons with lexical baselines expose differences between deployable fusion, oracle selection, and performance on label-poor targets.
网页跨平台
ComponentBench: Diagnosing Component-Level Failures in Computer-Use Agents
ComponentBench evaluates short interactions with canonical web UI components using verified tasks and matched human reference trajectories. Its diagnostic pipeline compares observation/action representations and analyzes structural difficulty, failure patterns and interaction efficiency.
网页
Improving LLM-based Autonomous Web Agents with Filtering
Trains HTML-element relevance filters on Mind2Web trajectories and transfers them to WebArena to reduce irrelevant browser context. It also evaluates a zero-shot ColBERT retriever and documents baseline web-agent failure modes.
网页
Hallucination-Free GUI Grounding via Regression-Free Layout-Aware Matching
This framework separates frozen-model instruction expansion from a trained layout-aware matcher that localizes GUI elements without coordinate regression. Binary text/icon supervision supports the matcher, with evaluations on GUI grounding and Mind2Web element-selection tasks.
网页
Not an A11y: How Android Accessibility Exposes Mobile AI Agents to Indirect Prompt Injection
Demonstrates that unsanitized Android accessibility tree metadata exposes mobile GUI agents to stealthy indirect prompt injection attacks.
Android
CAP: A Scalable Benchmark for Evaluating Cross-Site Browser Agents with Complex Actions and Perception
CAP builds cross-site browser workflows by decomposing websites into operation and perception requirements and recomposing them into tasks. Its judge-based evaluation diagnoses complex interactions and visual-perception failures across real websites.
网页
Routing Is Least Learnable Where It Is Most Valuable: Bounds on Representation Routing for Web Agents
Measures the complementarity, cost and rerun variability of six browser observation modes on WebArena and VisualWebArena. Its routing experiments distinguish an oracle cost opportunity from policies that fail to robustly outperform a fixed observation mode.
网页
GUI-Lens: Coarse-to-Fine Cropping for GUI Grounding with General-Purpose VLMs
GUI-Lens uses OCR and detected interface components as coordinate references while a general-purpose VLM selects progressively enlarged screenshot crops. Instruction-conditioned checks constrain the crop and click proposals before mapping the final location back to the original screen.
Desktop-Delta Bench: Do Computer-Use Models Understand Desktop GUI Transitions?
Introduces Desktop-Delta Bench, an offline diagnostic of GUI transition understanding using temporal-ordering tasks and before/after action inference. It includes decoy observations to test whether models track the source and task relevance of screen changes.
桌面Linux
Do GUI Agents Believe Their Eyes? Diagnosing State-Belief Reliance on Pixels versus Structure
Defines visual state reliance and probes whether GUI agents resolve screenshot/structure conflicts using pixels, serialized interface structure or prior beliefs. Paired interventions and live multi-step tests connect misattributed state beliefs to failures and compare consistency-based mitigations.
跨平台
LUMOS: A Semantic Operating-System Layer for Accessibility-Grounded AI Agents
LUMOS converts native accessibility metadata and browser UI structure into semantic interface blueprints with stable element identifiers and available actions. It combines live pointer-to-element grounding with an observe-act loop constrained to visible-UI primitives.
网页
One Forward Beats Two: InnerZoom for Accurate and Efficient GUI Grounding
InnerZoom preserves target-region cues from intermediate decoder layers and reinjects a compact evidence state during coordinate decoding. Its single-forward GUI grounder is compared with crop-and-rerun zooming and matched supervised/RL baselines for accuracy and inference cost.
GUICrafter: Weakly-Supervised GUI Agent Leveraging Massive Unannotated Screenshots
GUICrafter first learns grounding from large collections of unannotated screenshots and webpages, then uses a smaller high-quality dataset for reinforcement-learning calibration. The curriculum targets reduced dependence on manually annotated GUI training data.
Agent-Computer Observation Interfaces Enable Dynamic Computer Use
The Agent-Computer Observation Interface separates continuous perception from discrete computer actions using captured frames, audio transcription and persistent visual narration. DynaCU-Bench tests these components on dynamic browser tasks and static controls, with model-specific ablations showing that additional observations can also hurt.
网页
VISUALSKILL: Multimodal Skills for Computer-Use Agents
VISUALSKILL organizes application-specific text and figures into hierarchical skills retrieved on demand through an MCP tool. Skills combine authored documentation with live UI exploration, and matched multimodal/text-only comparisons test their utility for interface identification and workflow-state verification.
CAPED: Context-Aware Privacy Exposure Defense for Mobile GUI Agents
CAPED filters mobile screenshots before upload by combining task requirements, screen context and parsed UI elements to mask incidental private content. AndroidWorld and a controlled seeded privacy evaluation examine leakage reduction alongside task utility and prototype costs.
Android
Signal-Driven Observation for Long-Horizon Web Agents
Proposes Signal-Driven Observation: a separate call selects relevant DOM elements and selectors, and a detector requests new observations only after meaningful browser events or failures. The paper frames observation timing and compression as an architectural research problem for long-horizon web agents.
网页
DragOn: A Benchmark and Dataset for Drag-Based GUI Interactions
DragOn supplies drag-grounding training data and a held-out benchmark for highlighting text, selecting cells, resizing elements and moving sliders. It evaluates existing vision-language models and a model fine-tuned on the drag tasks.
Benchmarking Living-Screen-Native GUI Agents on Short-Video Platforms
LivingScreen evaluates agents in a browser-based short-video environment where observing content consumes time and affects subsequent actions. Its task suite scores accuracy together with information efficiency to diagnose over- and under-observation on continuously changing screens.
网页
STaR-KV: Spatio-Temporal Adaptive Re-weighting for KV Cache Compression in GUI Vision-Language Models
STaR-KV compresses GUI-agent KV caches by calibrating visual-token importance across attention subspaces and interaction time. Its training-free scoring combines spatial specialization, temporal redundancy and an adaptive temperature, evaluated at matched cache budgets.
MUIAnno: An Expert-Annotated Dataset and Evaluation Benchmark for Mobile UI Understanding
MUIAnno provides expert labels for elements in manually explored mobile-app screens, with a web annotation tool that emits structured JSON. It also supplies UI-element detection baselines for the dataset.
DRS-GUI: Dynamic Region Search for Training-Free GUI Grounding
DRS-GUI searches screenshot regions through focus, shift and scatter operations selected by a tree-search planner. A region-quality reward guides this training-free perceptual search before final GUI grounding.
ReVision: Scaling Computer-Use Agents via Temporal Visual Redundancy Reduction
Trains GUI agents with a learned selector that removes repeated visual patches across consecutive screenshots while preserving spatial structure. ReVision evaluates this history representation on OSWorld, WebTailBench and AgentNetBench.
桌面网页
What Happens Before Decoding? Prefill Determines GUI Grounding in VLMs
Re-Prefill adds an attention-guided second prefill pass that revisits candidate screen regions before coordinate generation. The study evaluates this training-free intervention across GUI-grounding models and benchmarks.
Region4Web: Rethinking Observation Space Granularity for Web Agents
Reorganizes webpage accessibility trees into functional regions through hierarchical decomposition and semantic abstraction. PageDigest supplies a persistent region-level observation to web agents, evaluated on WebArena.
网页
AutoFocus: Uncertainty-Aware Active Visual Search for GUI Grounding
Presents AutoFocus, a training-free grounding method that turns sampled coordinates and token uncertainty into spatial region proposals. Shape-aware crops and visual comparison select a target while retaining surrounding interface context.
A11y-Compressor: A Framework for Enhancing the Efficiency of GUI Agent Observations through Visual Context Reconstruction and Redundancy Reduction
A11y-Compressor transforms linearized accessibility trees through modal detection, redundancy reduction and semantic structuring. Its OSWorld evaluation measures both observation-token reduction and task success.
Beyond Screenshots: Evaluating VLMs' Understanding of UI Animations
AniMINT evaluates how VLMs perceive, explain and interpret UI animations using densely annotated videos. The study distinguishes basic motion recognition from higher-level animation meaning and probes motion, context and perceptual cues.
SnapGuard: Lightweight Prompt Injection Detection for Screenshot-Based Web Agents
SnapGuard detects screenshot-based prompt injection through complementary visual-stability and recovered textual cues. Its evaluation compares detection quality and inference cost across attack and benign-page settings, rather than establishing universal runtime prevention.
网页
AutoGUI-v2: A Comprehensive Multi-Modal GUI Functionality Understanding Benchmark
AutoGUI-v2 evaluates interface functionality, grounding and predicted interaction outcomes using hierarchical regions derived through VLM-human annotation. Its tasks span multiple operating systems and expose differences between functional localization, captioning and transition-logic understanding.
跨平台
Do LLMs Need to See Everything? A Benchmark and Study of Failures in LLM-driven Smartphone Automation using Screentext vs. Screenshots
DailyDroid compares text-only and text-plus-screenshot agents on everyday Android tasks across multiple apps and difficulty levels. Its failure analysis examines accessibility, observation modality and agent/application behavior.
Android
PrecisionCUA: Iterative Visual Refinement for Pixel-Precise Cursor Grounding in Code Editors
Studies iterative cursor grounding in dense code-editor interfaces, using visual feedback from earlier attempts to correct displacement. PrecisionCUA evaluates this closed-loop process across multiple model families on coding-interface tasks.
What's Missing in Screen-to-Action? Towards a UI-in-the-Loop Paradigm for Multimodal GUI Reasoning
Introduces UILoop, which inserts explicit UI-element localization, function and usage reasoning between screen observation and action. UI Comprehension-Bench evaluates these element-level capabilities with three metrics.
PixelPrune: Pixel-Level Adaptive Visual Token Reduction via Predictive Coding
PixelPrune removes pixel-redundant patches before the vision encoder, reducing both encoding and downstream language-model work. The training-free method evaluates lossless and controlled-lossy compression on document and GUI tasks.
相关研究
MolmoPoint: Better Pointing for VLMs with Grounding Tokens
MolmoPoint replaces textual coordinate generation with pointing tokens that select visual regions and refine them through subpatches. The general image/video method includes GUI-pointing evaluation on ScreenSpotPro alongside other visual grounding tasks.
相关研究
Where and How to Prune: An Empirical Study of Visual Token Pruning for GUI Agent Navigation
Studies visual token pruning for GUI navigation, separating reusable screenshot encoder embeddings from later language-model computation. The analysis examines foreground-background token allocation and spatial coverage when pruning historical screenshots.
GUIDE: A Benchmark for Understanding and Assisting Users in Open-Ended GUI Tasks
GUIDE evaluates whether models can recognize a user’s behavior, infer intent and decide when to offer help during open-ended software tasks. It uses narrated screen recordings of novice users and studies the effect of additional user context on these assistance decisions.
MANA: Towards Efficient Mobile Ad Detection via Multimodal Agentic UI Navigation
MANA combines visual, temporal and exploration signals to guide navigation toward mobile advertisements. It evaluates ad detection and interaction cost on commercial smartphones, including obfuscated-ad cases.
AdaZoom-GUI: Adaptive Zoom-based GUI Grounding with Instruction Refinement
Combines instruction rewriting with conditional second-stage zooming for small GUI targets. A grounding dataset and GRPO training support joint prediction of click coordinates and element bounding boxes.
WebPII: Benchmarking Visual PII Detection for Computer-Use Agents
WebPII provides synthetic e-commerce screenshots with fine-grained annotations for sensitive information, including partially filled forms and transaction identifiers. It also trains WebRedact to detect visual PII as a privacy component for computer-use agents.
网页
GUI-CEval: A Hierarchical and Comprehensive Chinese Benchmark for Mobile GUI Agents
GUI-CEval evaluates Chinese mobile GUI agents through a hierarchy of atomic abilities and application-level tasks collected on physical devices. Its diagnostics cover perception, planning, reflection, execution and post-action evaluation across a broad set of apps.
Zoom to Essence: Trainless GUI Grounding by Inferring upon Interface Elements
Presents ZoomUI, a training-free grounding method that expresses instructions as visual element descriptions and uses model attention to zoom progressively into relevant interface regions. Evaluation examines instruction-to-element localization on GUI benchmarks.
PIRA-Bench: A Transition from Reactive GUI Agents to GUI-based Proactive Intent Recommendation Agents
PIRA-Bench evaluates proactive intent recommendations from noisy GUI screenshot sequences containing interleaved tasks and user context. Its PIRF baseline combines memory and state tracking to maintain multiple task threads and identify opportunities to assist.
第 1–60 条，共 212 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。