已审阅并标注「动作预测」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。
Separates next-screen prediction from action evaluation in GUI lookahead and trains Direct Action Ranking on reference actions, comparing prediction-augmented scoring under matched offline supervision.
Evaluates Jev as a bounded action selector inside three AndroidWorld agents and two text-based ALFWorld agents, separating decision savings from task success, recovery and resources per delivered outcome.
ShareMem consolidates reusable experience across users while keeping preference retrieval user-bound; its web evaluation predicts target elements and actions on recorded Mind2Web states.
WebMRE pairs guidance sentences with grounded actions in an offline benchmark derived from successful WebArena traces. Controlled decoding-prefix interventions test how guide content affects action selection, while separating offline prediction metrics from live browser task success.
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.
Introduces LACL-GUI, a contrastive learning objective that distinguishes trajectory quality within both successful and failed GUI executions. It favors concise successes and uses divergence from successful trajectories to differentiate failures during policy optimization.
Introduces State Transition Pretraining, which jointly learns to infer actions from GUI changes and predict future states from current screens and actions. Subsequent instruction-trajectory fine-tuning tests whether these representations transfer to desktop and mobile agent tasks.
Introduces EvoGUI, which derives temporal ordering, inverse action/value prediction, and successor-discrimination probes from recorded GUI trajectories. Its web-derived benchmark isolates transition understanding from the other abilities involved in end-to-end task success.
MAG jointly evaluates browser task actions and user-facing guides grounded in screenshots through element marks or pixel coordinates. Its harness covers annotation, live evaluation and joint metrics, with an expert-trajectory-augmented GRPO training study.
Introduces Uni-GUI, a collection of executable desktop and mobile trajectories, and UI-MOPD, which trains a shared policy from its own rollouts using platform-specific teachers. The distillation procedure routes each visited state to the appropriate teacher to preserve distinct interaction conventions.
Introduces CUActSpot to evaluate clicks, drags, drawing and other screen interactions across GUI, text, tables, canvases and natural images. A renderer-based pipeline records geometry and screenshots, then generates matching instructions and action traces for training a grounding model.
OmniGUI evaluates step-level action prediction from interleaved smartphone screenshots, synchronous audio and video. Expert demonstrations carry multimodal-dependency annotations, and initial baselines use native multimodal models as agent proxies to study temporal and audio-related failures.
Adapts a discrete vision-language diffusion model to single-turn GUI action and bounding-box prediction, using a hybrid masking schedule for coordinate geometry. Experiments across web, desktop and mobile datasets examine grounding accuracy and inference tradeoffs.
Introduces HATS, which directs GUI trajectory exploration toward semantically ambiguous actions and iteratively checks instruction-execution alignment. Feedback from trajectory refinement updates the exploration signal used to collect subsequent training examples.
CowCorpus records interleaved human and agent actions during web navigation and identifies distinct patterns of user intervention. Models trained to anticipate these interventions are integrated into live web agents and assessed in a user study.
Presents a desktop world model that predicts textual UI changes before synthesizing the next screenshot. It learns from Microsoft Office transitions and is evaluated by letting a frozen agent compare simulated actions before execution.
Introduces SwipeGen for synthesizing varied swipe trajectories and SwipeBench for assessing their execution quality. The study examines how richer control of trajectory, speed, and timing changes GUI agents’ interaction performance.
GUI-Cursor treats GUI grounding as a sequence of cursor movements guided by visible feedback and movement history. Online reinforcement learning with trajectory rewards trains the model to refine the target location across multiple steps.
UIPro pretrains GUI understanding on 20.6 million tasks and then harmonizes interaction datasets through a unified action space. Continued fine-tuning uses the merged multi-platform data to train action prediction for generalist GUI operation.
Introduces OmniActor, which shares shallow model layers while separating deeper parameters to train jointly on GUI and embodied interactions. A unified action representation and mixed training data support evaluation in both task families.
Introduces UITron-Speech, which predicts GUI actions from spoken instructions and screenshots. It combines synthetic speech data and mixed-modality training with a training-free grounding refinement step.
LiMAC combines a compact Action Transformer with a fine-tuned vision-language model for Android app control. It uses textual goals and histories of screenshots and UI trees to predict actions, with evaluation on two mobile-control datasets.
MobileVLM adds two pretraining stages and four UI tasks to learn element detail and relationships between successive screens. Its Mobile3M dataset contains three million Chinese mobile UI pages connected by recorded transition actions.
Describes V-Zen, a multimodal GUI model with dual-resolution image encoders for grounding and next-action prediction, fine-tuned using GUIDE interface elements and task sequences.
Trains a smartphone assistant from sequences of screenshots paired with earlier actions. The model predicts UI-only interactions such as taps and swipes, with evaluation on Android in the Wild.
Studies instruction-following agents that observe screenshots and act through a generic mouse-and-keyboard action space, evaluating pixel-based pretraining for GUI tasks on MiniWoB++.
Trains WebGUM offline from demonstrations by jointly tuning an instruction-following language model and a vision encoder over webpage screenshots and HTML. The agent predicts browser actions and is evaluated on MiniWoB, WebShop, and Mind2Web.
Pretrains ActionBert on user interaction traces to combine visual, linguistic, and UI-structure features. It evaluates the resulting UI representations on tasks including icon classification and description-based component retrieval.
Introduces PIXELHELP for predicting mobile UI action sequences from instructions, together with action-phrase annotations and synthetic UI action descriptions for training. Separate Transformers extract instruction steps and ground object descriptions using interface content and screen position.
Drag-and-click episode data for the ShowUI-π model, spanning desktop editing, presentation, handwriting and other interface tasks.
The released mobile training subset for the original UI-R1 3B model, with action annotations and individual screenshot files.
The context-aware Aria-UI release that uses interaction context to support grounding in dynamic Android and desktop agent tasks.
UI-TARS-1.5-7B-derived checkpoint trained with ARPO on OSWorld desktop tasks.
A bilingual GLM-4V-based execution model that uses screenshots, task instructions and action history to predict GUI operations.
A Qwen2.5-VL-based GUI checkpoint released with Mobile-Agent-v3 for grounded actions and specialized roles in its agent framework.
A single Hub release containing GUI-R1 3B and 7B action-model checkpoints trained with reinforcement learning across interface platforms.
A Qwen3-VL-based checkpoint for navigation and multi-step GUI actions across web, desktop and mobile interfaces.
A sparse Qwen3.5-based GUI-action checkpoint released for navigation and computer-use workflows.
A Qwen3.6-based computer-use checkpoint designed to request GUI actions, code execution and tool calls through an external agent harness.
A Llama 3 8B instruction-model checkpoint fine-tuned on WebLINX web interactions and dialogue, predicting browser actions from a textual state representation.
A Qwen2.5-VL-based computer-use checkpoint that combines screenshot history and reflective reasoning to predict executable desktop actions.
The AndroidControl-oriented Qwen2-VL checkpoint trained with mobile trajectories generated through OS-Genesis reverse task synthesis.
Computer-use checkpoint fine-tuned on PC-Agent-E trajectories; the model-card prose identifies Qwen2.5-VL-72B-Instruct as its starting model.
A Qwen2.5-VL-based computer-use checkpoint with direct grounding and agent action modes across desktop, mobile and web interfaces.
A compact visual-action checkpoint that predicts continuous GUI drag trajectories from a screenshot and a natural-language instruction.
A Qwen2.5-VL mobile action model trained using existing and reward-guided synthetic trajectories from the UI-Genie framework.
A Qwen2.5-VL checkpoint trained with reinforcement learning for GUI action prediction and evaluated on mobile, desktop and web grounding tasks.
A mobile GUI checkpoint trained with semi-online reinforcement learning that uses offline trajectories to improve multi-turn action prediction.
Research scripts that generate screen descriptions, action reasoning and expected outcomes, then predict Android GUI actions using the resulting Chain-of-Action-Thought context.
Offline prompting and evaluation code for Dynamic Planning of Thoughts on Android-in-the-Wild, including action extraction and comparison with reference actions.
A shared evaluation harness for GUI-model perception, offline action prediction and multi-step interaction, using simple smolagents-based agents across existing benchmarks.
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。
