已审阅并标注「状态追踪」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。
ECHOES learns recursive textual task memory through privileged screenshot-history distillation and GRPO, then predicts GUI actions from the current screenshot and memory.
LongPuzzleBench evaluates coherent long-horizon planning on persistent visual puzzle boards through native GUI actions. It compares GUI-only and code-assisted agents and uses controlled hints and memory diagnostics to examine decisions that preserve or eliminate future solutions.
Introduces StateAliasBench to isolate GUI world-model ambiguities in which matching visible screens and actions can lead to different futures. It augments frozen world models with structured state recovered from interaction history and evaluates the effect on prediction and agent execution.
MaP unifies GUI-navigation objectives by masking and predicting components of multi-turn trajectories. Role-aware adapters address heterogeneous training data, and evaluations compare the approach with direct task-mixture training across navigation benchmarks.
BreakingWeb pairs solvable browser tasks with controlled environment interventions that preserve the goal, target and backend success criterion. The paired evaluations isolate robustness failures, including agents declaring completion without the required state change.
AnchorGUI compares expected and observed UI transitions and selectively retains screenshots when they disagree. This asymmetric memory supports immediate correction and focuses cross-attempt experience distillation on likely failure steps.
JarvisGUI composes typed input-output task transformations into dynamic workflows spanning Android, Windows and Ubuntu devices. Its virtual-environment benchmark tests transfer of intermediate results, shared state and cross-device dependencies.
Introduces LocalLSTC, a training-free control architecture that maintains persistent task state and uses it to form short execution commitments. Outcomes feed back into that state for progress checks, recovery, and termination in locally deployed desktop agents.
SynWeaver explores a website to build a map of page states and executable interactions, then learns website priors for task proposals. Joint task/trajectory synthesis revises inconsistent pairs and verifies or repairs the resulting supervision.
StepReflect predicts structured per-action reflections from transition specifications and paired visual evidence. The paper distinguishes offline transition-label accuracy from online task success when the reflection model is integrated into mobile agents.
CUADebug combines a failure taxonomy and human-annotated OSWorld failure trajectories with an agent that locates causal steps and proposes corrections from screenshots and action traces. It evaluates root-cause diagnosis separately from controlled task re-execution and failure recovery.
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.
Introduces StateAct, a code-first computer-use harness that manipulates program state directly and delegates screenshot-and-click subgoals to a GUI agent. An independent completion gate checks saved outputs, while fresh subagents handle subgoals to limit accumulated context.
StructAgent maintains compact task state and regulates long-horizon execution through verified transitions and progress checkpoints. Its workflow supports evidence-based completion and targeted recovery, with desktop computer-use evaluation and a separate Minecraft extension.
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.
VisCritic compares before/after screenshots through change-aware visual features and an action-conditioned critic. Weakly supervised samples from existing traces train judgments of action success, progress and error type for GUI-agent verification.
ChainWorld composes compatible atomic OSWorld tasks into multi-objective desktop chains while preserving their evaluators. It compares giving the full sequence at once with revealing goals turn by turn, exposing different state and session-management failures.
StainFlow tracks visually verifiable task entities through GUI trajectories to identify phases and retrieve focused evidence windows. Its process-reward model is tested for both online policy learning and trajectory-completion judgments.
MACU represents computer-use objectives as a dependency graph whose manager dispatches parallel subagents and revises tasks as new information arrives. Desktop and web experiments examine coordination, information handoff and task-completion time.
UI-KOBE explores mobile applications to build graphs of UI states and executable transitions. A lightweight agent then matches its current screenshot to the graph and chooses a transition, local action, completion signal or fallback action.
SaaS-Bench evaluates browser agents on professional workflows across self-hosted applications. Weighted checkpoints measure task completion and partial progress, with application-state checks and configured content judges documented in the official harness.
LongMemEval-V2 asks memory systems to recover environment knowledge from long web-agent histories for question answering. Its AgentRunbook variants contrast indexed knowledge pools with coding-agent retrieval from stored trajectory files.
Proposes a step-level model cascade that runs a smaller computer-use policy until learned monitors flag stalled progress or semantic drift. Separate history-based stuck detection and milestone checks determine when to escalate for recovery or verification without retraining the larger model.
Introduces DynamicGUIBench for applications whose interfaces change between actions, together with DynamicUI, an agent that processes screen recordings. The agent selects informative frames, filters inconsistent action reasoning and uses trajectory reflection for subsequent decisions.
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.
This study examines UI changes between a desktop agent’s screenshot observation and action dispatch. Its pre-execution state checks address overlay and focus manipulation, while the reported DOM-injection blind spot limits the defense’s coverage.
AJ-Bench evaluates judge agents that actively gather evidence from search, data systems and graphical interfaces. Its annotated tasks and trajectories test information acquisition, state verification and process verification.
WebSP-Eval evaluates agents carrying out website security and privacy tasks such as preference changes and session management. A browser extension controls account and initial state across runs, enabling analysis of failures on stateful UI controls.
Introduces VeriGUI, which checks expected action effects and reasons about recovery before subsequent GUI actions. Training combines synthetic failure trajectories with verification-weighted GRPO, evaluated on an AndroidControl-based robustness benchmark.
GUIDE evaluates GUI trajectories by segmenting them into subtasks, diagnosing each segment, and aggregating structured findings into a task verdict. It tests diagnostic judgments on e-commerce, web-agent and Android-control traces rather than relying on one holistic trajectory score.
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.
Identifies time-of-check to time-of-use (TOCTOU) race vulnerabilities when web pages mutate between agent planning and action execution.
UI-Oceanus converts autonomous GUI exploration and execution feedback into supervision for predicting future interface states. It compares forward and inverse objectives during continual pretraining and examines downstream offline and online navigation as synthetic data scales.
BEAP-Agent treats GUI execution as depth-first search with coordinated planning, execution and task-state tracking. It supports backtracking across multiple levels of earlier states and evaluates recovery on OSWorld.
DrawingBench evaluates sequences of low-level mouse-based GUI actions through 250 drawing prompts and rule-based criteria. Action inspection and multi-turn external feedback support analysis of spatial reasoning, tool-state management, and refinement.
This work trains high-level schedulers with staged reinforcement learning driven by execution feedback. Its CES framework separates a Coordinator for planning and decomposition, a StateTracker for compressed context, and a low-level Executor.
Introduces MGA, which separates task-agnostic screen-state reading from decision making and connects steps through structured state memory. Validated state deltas replace full trajectory replay, evaluated on OSWorld and application tasks.
ColorBench represents mobile tasks as graphs of states collected from real devices, providing a quasi-dynamic environment with multiple valid paths and error paths. It evaluates single-app and cross-app tasks using completion, subtask progress, and atomic diagnostics.
Introduces a benchmark for binary GUI-toggle instructions and State-aware Reasoning, which compares the observed control state with the requested state before choosing an action. It evaluates agents on toggle tasks and broader interactive benchmarks.
Chain-of-Memory explicitly records action descriptions and task-relevant screen information in short- and long-term memory for GUI navigation. GUI Odyssey-CoM provides 111,000 screen-action pairs with memory annotations for learning and evaluating information retention across applications.
Analyzes how interaction history and webpage representation affect multiturn web-agent generalization across unfamiliar websites, task categories, and geographic settings.
OSCAR translates user instructions into Python control code while observing screen images and acting through mouse and keyboard operations. A state-machine controller handles errors and replans tasks during execution across desktop and mobile applications.
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.
Trident combines exploration, monitoring, and detection agents to identify functional bugs that do not crash a mobile app. It reasons over screenshot sequences and interface transitions to construct test oracles, evaluated on 590 non-crash bugs and additional live apps.
Studies explicit textual estimates of latent UI state, including whether commanded actions occurred and how far a task has progressed. The experiments compare agents that reason with these estimates against agents that omit them across public and internal task benchmarks.
LlamaTouch executes mobile tasks on devices and evaluates whether an agent visits manually annotated essential application or system states. Screenshot and hierarchy annotations, combined with exact and fuzzy state matching, support evaluation of 496 tasks across multiple agents.
Android GUI-agent framework that keeps long-running task context in structured programs, variables and control flow, with a CLI and AndroidWorld integration.
State-tracking prompts and experiment runners for language agents, including a WebShop implementation alongside text-based environments.
Runtime verification framework for Android GUI agents that checks proposed actions against logical rules derived from user instructions and developer-defined app states.
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。
