{ "version": "https://jsonfeed.org/version/1.1", "title": "AwesomeCUA: newly reviewed resources", "description": "Computer-use agent papers, models, datasets, benchmarks and tools as they pass source review, newest review first. Metadata-only bibliography records and unreviewed candidates are not included.", "home_page_url": "https://awesomecua.com/", "feed_url": "https://awesomecua.com/feed.json", "language": "en", "favicon": "https://awesomecua.com/favicon.svg", "authors": [ { "name": "AwesomeCUA", "url": "https://awesomecua.com/" } ], "items": [ { "id": "https://awesomecua.com/resources/arxiv-2510-04536/", "url": "https://awesomecua.com/resources/arxiv-2510-04536/", "external_url": "https://arxiv.org/abs/2510.04536", "title": "3Dify: a Framework for Procedural 3D-CG Generation Assisted by LLMs Using MCP and RAG", "content_text": "Introduces 3Dify, a natural-language 3D graphics framework that operates creation tools through MCP and uses a computer-using agent for GUI operations when MCP is unavailable.", "summary": "Introduces 3Dify, a natural-language 3D graphics framework that operates creation tools through MCP and uses a computer-using agent for GUI operations when MCP is unavailable.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "automation", "tool-integration", "science" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "unspecified" ], "record_json": "https://awesomecua.com/resources/arxiv-2510-04536/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2510-04536/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2510-06223/", "url": "https://awesomecua.com/resources/arxiv-2510-06223/", "external_url": "https://arxiv.org/abs/2510.06223", "title": "A Multimodal GUI Architecture for Interfacing with LLM-Based Conversational Assistants", "content_text": "Presents an MCP architecture that exposes an application's GUI navigation graph and view-specific tools to speech-enabled LLM assistants for voice interaction.", "summary": "Presents an MCP architecture that exposes an application's GUI navigation graph and view-specific tools to speech-enabled LLM assistants for voice interaction.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "tool-integration", "ui-design" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "unspecified" ], "record_json": "https://awesomecua.com/resources/arxiv-2510-06223/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2510-06223/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2602-14364/", "url": "https://awesomecua.com/resources/arxiv-2602-14364/", "external_url": "https://arxiv.org/abs/2602.14364", "title": "A Trajectory-Based Safety Audit of Clawdbot (OpenClaw)", "content_text": "Audits Clawdbot safety through logged action and tool-call trajectories across local execution and web-mediated workflows, using automated and human review.", "summary": "Audits Clawdbot safety through logged action and tool-call trajectories across local execution and web-mediated workflows, using automated and human review.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "safety-security", "evaluation", "observability", "workflow" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "web" ], "record_json": "https://awesomecua.com/resources/arxiv-2602-14364/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2602-14364/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2510-24702/", "url": "https://awesomecua.com/resources/arxiv-2510-24702/", "external_url": "https://arxiv.org/abs/2510.24702", "title": "Agent Data Protocol: Unifying Datasets for Diverse, Effective Fine-tuning of LLM Agents", "content_text": "The agent data protocol (ADP) provides a shared format for agent training data across browsing, tool use, coding, and other tasks, including web agent workflows.", "summary": "The agent data protocol (ADP) provides a shared format for agent training data across browsing, tool use, coding, and other tasks, including web agent workflows.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "data-generation", "workflow" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "unspecified" ], "record_json": "https://awesomecua.com/resources/arxiv-2510-24702/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2510-24702/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2601-03641/", "url": "https://awesomecua.com/resources/arxiv-2601-03641/", "external_url": "https://arxiv.org/abs/2601.03641", "title": "Agent-Dice: Disentangling Knowledge Updates via Geometric Consensus for Agent Continual Learning", "content_text": "Agent-Dice fuses parameter updates to support continual learning in LLM agents, with GUI agent tasks evaluated alongside tool-use domains.", "summary": "Agent-Dice fuses parameter updates to support continual learning in LLM agents, with GUI agent tasks evaluated alongside tool-use domains.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "robustness", "efficiency" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "unspecified" ], "record_json": "https://awesomecua.com/resources/arxiv-2601-03641/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2601-03641/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2601-12560/", "url": "https://awesomecua.com/resources/arxiv-2601-12560/", "external_url": "https://arxiv.org/abs/2601.12560", "title": "Agentic Artificial Intelligence (AI): Architectures, Taxonomies, and Evaluation of Large Language Model Agents", "content_text": "Surveys LLM agent architectures, operating environments and evaluation, covering web navigation and native computer use alongside other agent domains.", "summary": "Surveys LLM agent architectures, operating environments and evaluation, covering web navigation and native computer use alongside other agent domains.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "survey", "survey", "evaluation", "planning", "tool-integration" ], "_awesomecua": { "kind": "survey", "review_status": "curated", "environments": [ "web" ], "record_json": "https://awesomecua.com/resources/arxiv-2601-12560/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2601-12560/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2510-14545/", "url": "https://awesomecua.com/resources/arxiv-2510-14545/", "external_url": "https://arxiv.org/abs/2510.14545", "title": "Agentic Entropy-Balanced Policy Optimization", "content_text": "Introduces Agentic Entropy-Balanced Policy Optimization (AEPO), an RL method that adjusts rollout sampling and policy updates for tool-using web agents.", "summary": "Introduces Agentic Entropy-Balanced Policy Optimization (AEPO), an RL method that adjusts rollout sampling and policy updates for tool-using web agents.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "reinforcement-learning", "efficiency" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "web" ], "record_json": "https://awesomecua.com/resources/arxiv-2510-14545/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2510-14545/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2509-11067/", "url": "https://awesomecua.com/resources/arxiv-2509-11067/", "external_url": "https://arxiv.org/abs/2509.11067", "title": "Agentic Lybic: Multi-Agent Execution System with Tiered Reasoning and Orchestration", "content_text": "Introduces Agentic Lybic, a desktop automation system that routes tasks among GUI, code, and decision-support workers using a finite-state machine and evaluator.", "summary": "Introduces Agentic Lybic, a desktop automation system that routes tasks among GUI, code, and decision-support workers using a finite-state machine and evaluator.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "automation", "planning", "tool-integration", "workflow" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "desktop" ], "record_json": "https://awesomecua.com/resources/arxiv-2509-11067/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2509-11067/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2509-19199/", "url": "https://awesomecua.com/resources/arxiv-2509-19199/", "external_url": "https://arxiv.org/abs/2509.19199", "title": "Agentic Reinforcement Learning with Implicit Step Rewards", "content_text": "Introduces iStar, a general agentic RL method that learns implicit step rewards; WebShop is one of several interactive benchmarks used to evaluate it.", "summary": "Introduces iStar, a general agentic RL method that learns implicit step rewards; WebShop is one of several interactive benchmarks used to evaluate it.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "reinforcement-learning", "reward-models", "evaluation" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "unspecified" ], "record_json": "https://awesomecua.com/resources/arxiv-2509-19199/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2509-19199/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2511-08325/", "url": "https://awesomecua.com/resources/arxiv-2511-08325/", "external_url": "https://arxiv.org/abs/2511.08325", "title": "AgentPRM: Process Reward Models for LLM Agents via Step-Wise Promise and Progress", "content_text": "AgentPRM scores sequential agent decisions by their promise and progress toward a goal, with web shopping and browser navigation among the motivating tasks.", "summary": "AgentPRM scores sequential agent decisions by their promise and progress toward a goal, with web shopping and browser navigation among the motivating tasks.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "reward-models", "planning", "web-navigation" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "web" ], "record_json": "https://awesomecua.com/resources/arxiv-2511-08325/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2511-08325/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2510-10049/", "url": "https://awesomecua.com/resources/arxiv-2510-10049/", "external_url": "https://arxiv.org/abs/2510.10049", "title": "ALLOY: Generating Reusable Agent Workflows from User Demonstration", "content_text": "Introduces ALLOY, which turns user demonstrations into editable, reusable workflows for LLM-based web agents carrying out preference-dependent tasks.", "summary": "Introduces ALLOY, which turns user demonstrations into editable, reusable workflows for LLM-based web agents carrying out preference-dependent tasks.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "human-in-the-loop", "workflow", "imitation-learning" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "web" ], "record_json": "https://awesomecua.com/resources/arxiv-2510-10049/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2510-10049/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2511-19957/", "url": "https://awesomecua.com/resources/arxiv-2511-19957/", "external_url": "https://arxiv.org/abs/2511.19957", "title": "AppSelectBench: Application-Level Tool Selection Benchmark", "content_text": "AppSelectBench evaluates whether computer-using agents choose the right desktop application for a task, using generated user intents and multiple evaluation protocols.", "summary": "AppSelectBench evaluates whether computer-using agents choose the right desktop application for a task, using generated user intents and multiple evaluation protocols.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "evaluation", "planning", "data-generation" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "desktop" ], "record_json": "https://awesomecua.com/resources/arxiv-2511-19957/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2511-19957/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2509-04198/", "url": "https://awesomecua.com/resources/arxiv-2509-04198/", "external_url": "https://arxiv.org/abs/2509.04198", "title": "Are LLM Agents the New RPA? A Comparative Study with RPA Across Enterprise Workflows", "content_text": "Compares an agent using computer interfaces with UiPath RPA on enterprise data entry, monitoring, and document extraction, measuring speed, reliability, and development effort.", "summary": "Compares an agent using computer interfaces with UiPath RPA on enterprise data entry, monitoring, and document extraction, measuring speed, reliability, and development effort.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "evaluation", "automation", "industry", "workflow" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "unspecified" ], "record_json": "https://awesomecua.com/resources/arxiv-2509-04198/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2509-04198/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2601-06031/", "url": "https://awesomecua.com/resources/arxiv-2601-06031/", "external_url": "https://arxiv.org/abs/2601.06031", "title": "Beyond Clicking:A Step Towards Generalist GUI Grounding via Text Dragging", "content_text": "GUI-Drag provides synthesized text-dragging examples, while ScreenDrag benchmarks selecting text by dragging across varied GUI contexts.", "summary": "GUI-Drag provides synthesized text-dragging examples, while ScreenDrag benchmarks selecting text by dragging across varied GUI contexts.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "grounding", "data-generation", "evaluation" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "unspecified" ], "record_json": "https://awesomecua.com/resources/arxiv-2601-06031/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2601-06031/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2510-16720/", "url": "https://awesomecua.com/resources/arxiv-2510-16720/", "external_url": "https://arxiv.org/abs/2510.16720", "title": "Beyond Pipelines: A Survey of the Paradigm Shift toward Model-Native Agentic AI", "content_text": "Surveys model-native agentic AI and the role of RL in learned planning, tool use and memory, covering GUI agents as one application alongside deep-research agents.", "summary": "Surveys model-native agentic AI and the role of RL in learned planning, tool use and memory, covering GUI agents as one application alongside deep-research agents.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "survey", "survey", "reinforcement-learning", "planning" ], "_awesomecua": { "kind": "survey", "review_status": "curated", "environments": [ "unspecified" ], "record_json": "https://awesomecua.com/resources/arxiv-2510-16720/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2510-16720/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2602-10915/", "url": "https://awesomecua.com/resources/arxiv-2602-10915/", "external_url": "https://arxiv.org/abs/2602.10915", "title": "Blind Gods and Broken Screens: Architecting a Secure, Intent-Centric Mobile Agent Operating System", "content_text": "Analyzes security risks in screen-based mobile agents and proposes Aura, an agent runtime with structured interactions, app-agent isolation and mediated communication.", "summary": "Analyzes security risks in screen-based mobile agents and proposes Aura, an agent runtime with structured interactions, app-agent isolation and mediated communication.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "safety-security", "mobile-navigation", "sandbox" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "unspecified" ], "record_json": "https://awesomecua.com/resources/arxiv-2602-10915/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2602-10915/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2510-19838/", "url": "https://awesomecua.com/resources/arxiv-2510-19838/", "external_url": "https://arxiv.org/abs/2510.19838", "title": "Branch-and-Browse: Efficient and Controllable Web Exploration with Tree-Structured Reasoning and Action Memory", "content_text": "Branch-and-Browse is a web agent framework that uses tree-structured subtasks, web state replay, and page action memory to explore and complete tasks.", "summary": "Branch-and-Browse is a web agent framework that uses tree-structured subtasks, web state replay, and page action memory to explore and complete tasks.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "planning", "memory", "efficiency", "web-navigation" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "web" ], "record_json": "https://awesomecua.com/resources/arxiv-2510-19838/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2510-19838/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2511-20597/", "url": "https://awesomecua.com/resources/arxiv-2511-20597/", "external_url": "https://arxiv.org/abs/2511.20597", "title": "BrowseSafe: Understanding and Preventing Prompt Injection Within AI Browser Agents", "content_text": "BrowseSafe builds a benchmark of prompt injections in realistic HTML for browser agents and evaluates defenses against attacks that can affect agent actions.", "summary": "BrowseSafe builds a benchmark of prompt injections in realistic HTML for browser agents and evaluates defenses against attacks that can affect agent actions.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "safety-security", "evaluation", "web-navigation" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "web" ], "record_json": "https://awesomecua.com/resources/arxiv-2511-20597/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2511-20597/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2511-13725/", "url": "https://awesomecua.com/resources/arxiv-2511-13725/", "external_url": "https://arxiv.org/abs/2511.13725", "title": "Can We Stop Malicious AI? KILLBENCH: A Benchmark for External AI Kill Switch Feasibility", "content_text": "KillBench evaluates external prompt-style signals intended to halt maliciously operating web agents, across harmful scenarios and agent configurations.", "summary": "KillBench evaluates external prompt-style signals intended to halt maliciously operating web agents, across harmful scenarios and agent configurations.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "evaluation", "safety-security", "web-navigation" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "web" ], "record_json": "https://awesomecua.com/resources/arxiv-2511-13725/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2511-13725/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2510-19386/", "url": "https://awesomecua.com/resources/arxiv-2510-19386/", "external_url": "https://arxiv.org/abs/2510.19386", "title": "ColorAgent: Building A Robust, Personalized, and Interactive OS Agent", "content_text": "Presents ColorAgent, an operating system agent for long-horizon user instructions, trained with step-wise RL and self-evolving training and evaluated on AndroidWorld and AndroidLab.", "summary": "Presents ColorAgent, an operating system agent for long-horizon user instructions, trained with step-wise RL and self-evolving training and evaluated on AndroidWorld and AndroidLab.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "long-horizon", "reinforcement-learning", "human-in-the-loop" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "unspecified" ], "record_json": "https://awesomecua.com/resources/arxiv-2510-19386/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2510-19386/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2601-16206/", "url": "https://awesomecua.com/resources/arxiv-2601-16206/", "external_url": "https://arxiv.org/abs/2601.16206", "title": "Computer Environments Elicit General Agentic Intelligence in LLMs", "content_text": "LLM-in-Sandbox gives agents a minimal computer environment for resource access, file management and code execution; LLM-in-Sandbox-RL trains models to use it.", "summary": "LLM-in-Sandbox gives agents a minimal computer environment for resource access, file management and code execution; LLM-in-Sandbox-RL trains models to use it.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "sandbox", "tool-integration", "reinforcement-learning", "efficiency" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "unspecified" ], "record_json": "https://awesomecua.com/resources/arxiv-2601-16206/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2601-16206/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2510-03612/", "url": "https://awesomecua.com/resources/arxiv-2510-03612/", "external_url": "https://arxiv.org/abs/2510.03612", "title": "Cross-Modal Content Optimization for Steering Web Agent Preferences", "content_text": "Introduces Cross-Modal Preference Steering (CPS), a black-box attack that edits a listing's images and text to steer VLM-based web agents' selections.", "summary": "Introduces Cross-Modal Preference Steering (CPS), a black-box attack that edits a listing's images and text to steer VLM-based web agents' selections.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "safety-security", "web-navigation" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "web" ], "record_json": "https://awesomecua.com/resources/arxiv-2510-03612/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2510-03612/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2510-18596/", "url": "https://awesomecua.com/resources/arxiv-2510-18596/", "external_url": "https://arxiv.org/abs/2510.18596", "title": "CUARewardBench: A Benchmark for Evaluating Reward Models on Computer-using Agent", "content_text": "Introduces CUARewardBench for evaluating outcome and process reward models on annotated computer-use trajectories, plus Unanimous Prompt Ensemble (UPE) for reward judgments.", "summary": "Introduces CUARewardBench for evaluating outcome and process reward models on annotated computer-use trajectories, plus Unanimous Prompt Ensemble (UPE) for reward judgments.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "evaluation", "reward-models" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "unspecified" ], "record_json": "https://awesomecua.com/resources/arxiv-2510-18596/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2510-18596/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2509-21799/", "url": "https://awesomecua.com/resources/arxiv-2509-21799/", "external_url": "https://arxiv.org/abs/2509.21799", "title": "D-Artemis: A Deliberative Cognitive Framework for Mobile GUI Multi-Agents", "content_text": "Introduces D-Artemis, a mobile GUI agent framework that retrieves app-specific tips, checks actions before execution, and reflects on their outcomes without trajectory training.", "summary": "Introduces D-Artemis, a mobile GUI agent framework that retrieves app-specific tips, checks actions before execution, and reflects on their outcomes without trajectory training.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "mobile-navigation", "memory", "self-reflection", "training-free" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "android" ], "record_json": "https://awesomecua.com/resources/arxiv-2509-21799/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2509-21799/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2508-17398/", "url": "https://awesomecua.com/resources/arxiv-2508-17398/", "external_url": "https://arxiv.org/abs/2508.17398", "title": "DashboardQA: Benchmarking Multimodal Agents for Question Answering on Interactive Dashboards", "content_text": "Introduces DashboardQA to evaluate how vision-language GUI agents answer questions by interacting with dashboards, grounding elements, and planning actions.", "summary": "Introduces DashboardQA to evaluate how vision-language GUI agents answer questions by interacting with dashboards, grounding elements, and planning actions.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "evaluation", "grounding", "planning", "screen-understanding" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "unspecified" ], "record_json": "https://awesomecua.com/resources/arxiv-2508-17398/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2508-17398/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2512-22894/", "url": "https://awesomecua.com/resources/arxiv-2512-22894/", "external_url": "https://arxiv.org/abs/2512.22894", "title": "DECEPTICON: How Dark Patterns Manipulate Web Agents", "content_text": "DECEPTICON is an environment of 700 web navigation tasks that isolates dark patterns to measure how deceptive interface designs steer agent actions.", "summary": "DECEPTICON is an environment of 700 web navigation tasks that isolates dark patterns to measure how deceptive interface designs steer agent actions.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "evaluation", "ui-design", "safety-security", "web-navigation" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "web" ], "record_json": "https://awesomecua.com/resources/arxiv-2512-22894/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2512-22894/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2510-05137/", "url": "https://awesomecua.com/resources/arxiv-2510-05137/", "external_url": "https://arxiv.org/abs/2510.05137", "title": "Demystifying deep search: a holistic evaluation with hint-free multi-hop questions and factorised metrics", "content_text": "Introduces WebDetective, a deep-search benchmark with a traceable Wikipedia sandbox, and EvidenceLoop, an evidence-tracking workflow for web-agent search and synthesis.", "summary": "Introduces WebDetective, a deep-search benchmark with a traceable Wikipedia sandbox, and EvidenceLoop, an evidence-tracking workflow for web-agent search and synthesis.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "evaluation", "observability", "planning" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "web" ], "record_json": "https://awesomecua.com/resources/arxiv-2510-05137/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2510-05137/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2510-12798/", "url": "https://awesomecua.com/resources/arxiv-2510-12798/", "external_url": "https://arxiv.org/abs/2510.12798", "title": "Detect Anything via Next Point Prediction", "content_text": "Introduces Rex-Omni, a general visual perception model trained for coordinate prediction; GUI grounding is one capability evaluated alongside detection and OCR.", "summary": "Introduces Rex-Omni, a general visual perception model trained for coordinate prediction; GUI grounding is one capability evaluated alongside detection and OCR.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "grounding", "evaluation", "data-generation" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "unspecified" ], "record_json": "https://awesomecua.com/resources/arxiv-2510-12798/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2510-12798/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2602-07787/", "url": "https://awesomecua.com/resources/arxiv-2602-07787/", "external_url": "https://arxiv.org/abs/2602.07787", "title": "Do Multi-Agents Dream of Electric Screens? Achieving Perfect Accuracy on AndroidWorld Through Task Decomposition", "content_text": "Minitap coordinates six specialized agents for AndroidWorld tasks, checking text input against device state and detecting action loops to revise its strategy.", "summary": "Minitap coordinates six specialized agents for AndroidWorld tasks, checking text input against device state and detecting action loops to revise its strategy.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "mobile-navigation", "planning", "state-tracking", "self-reflection" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "android" ], "record_json": "https://awesomecua.com/resources/arxiv-2602-07787/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2602-07787/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2509-25301/", "url": "https://awesomecua.com/resources/arxiv-2509-25301/", "external_url": "https://arxiv.org/abs/2509.25301", "title": "Flash-Searcher: Fast and Effective Web Agents via DAG-Based Parallel Execution", "content_text": "Introduces Flash-Searcher, a parallel agent framework that organizes dependent subtasks as a graph and runs independent paths concurrently; it is evaluated on deep-search benchmarks.", "summary": "Introduces Flash-Searcher, a parallel agent framework that organizes dependent subtasks as a graph and runs independent paths concurrently; it is evaluated on deep-search benchmarks.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "efficiency", "planning", "workflow" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "unspecified" ], "record_json": "https://awesomecua.com/resources/arxiv-2509-25301/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2509-25301/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2510-03204/", "url": "https://awesomecua.com/resources/arxiv-2510-03204/", "external_url": "https://arxiv.org/abs/2510.03204", "title": "FocusAgent: Simple Yet Effective Ways of Trimming the Large Context of Web Agents", "content_text": "Introduces FocusAgent, which uses a lightweight LLM to select task-relevant lines from web agents' accessibility-tree observations, reducing context size and exposure to prompt injection.", "summary": "Introduces FocusAgent, which uses a lightweight LLM to select task-relevant lines from web agents' accessibility-tree observations, reducing context size and exposure to prompt injection.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "efficiency", "safety-security", "screen-understanding" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "web" ], "record_json": "https://awesomecua.com/resources/arxiv-2510-03204/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2510-03204/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2601-04203/", "url": "https://awesomecua.com/resources/arxiv-2601-04203/", "external_url": "https://arxiv.org/abs/2601.04203", "title": "FronTalk: Benchmarking Front-End Development as Conversational Code Generation with Multi-Modal Feedback", "content_text": "FronTalk benchmarks multi-turn front-end code generation from text and visual feedback; a web agent explores generated sites for evaluation and in AceCoder.", "summary": "FronTalk benchmarks multi-turn front-end code generation from text and visual feedback; a web agent explores generated sites for evaluation and in AceCoder.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "ui-design", "evaluation", "web-navigation" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "web" ], "record_json": "https://awesomecua.com/resources/arxiv-2601-04203/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2601-04203/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2507-23773/", "url": "https://awesomecua.com/resources/arxiv-2507-23773/", "external_url": "https://arxiv.org/abs/2507.23773", "title": "General Agentic Planning Through Simulative Reasoning with World Models", "content_text": "Introduces SiRA, an agent architecture that uses a language-based world model to simulate action outcomes for planning in web-browser tasks.", "summary": "Introduces SiRA, an agent architecture that uses a language-based world model to simulate action outcomes for planning in web-browser tasks.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "planning", "world-models", "web-navigation" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "web" ], "record_json": "https://awesomecua.com/resources/arxiv-2507-23773/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2507-23773/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2510-18314/", "url": "https://awesomecua.com/resources/arxiv-2510-18314/", "external_url": "https://arxiv.org/abs/2510.18314", "title": "Genesis: Evolving Attack Strategies for LLM Web Agent Red-Teaming", "content_text": "Introduces Genesis, a red-teaming framework that generates adversarial injections for web agents and revises attack strategies using responses and interaction logs.", "summary": "Introduces Genesis, a red-teaming framework that generates adversarial injections for web agents and revises attack strategies using responses and interaction logs.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "safety-security", "interaction-history" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "web" ], "record_json": "https://awesomecua.com/resources/arxiv-2510-18314/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2510-18314/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2510-20333/", "url": "https://awesomecua.com/resources/arxiv-2510-20333/", "external_url": "https://arxiv.org/abs/2510.20333", "title": "GhostEI-Bench: Do Mobile Agents Resilience to Environmental Injection in Dynamic On-Device Environments?", "content_text": "Introduces GhostEI-Bench, a benchmark that injects deceptive UI events into Android emulator workflows to assess mobile agents and analyze failures from actions and screenshots.", "summary": "Introduces GhostEI-Bench, a benchmark that injects deceptive UI events into Android emulator workflows to assess mobile agents and analyze failures from actions and screenshots.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "evaluation", "safety-security", "sandbox", "observability" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "android" ], "record_json": "https://awesomecua.com/resources/arxiv-2510-20333/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2510-20333/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2510-00507/", "url": "https://awesomecua.com/resources/arxiv-2510-00507/", "external_url": "https://arxiv.org/abs/2510.00507", "title": "Graph2Eval: Automatic Multimodal Task Generation for Agents via Knowledge Graphs", "content_text": "Introduces Graph2Eval, which generates multimodal agent evaluation tasks from knowledge graphs, including multi-step web interactions in Graph2Eval-Bench.", "summary": "Introduces Graph2Eval, which generates multimodal agent evaluation tasks from knowledge graphs, including multi-step web interactions in Graph2Eval-Bench.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "data-generation", "evaluation", "web-navigation" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "web" ], "record_json": "https://awesomecua.com/resources/arxiv-2510-00507/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2510-00507/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2510-26098/", "url": "https://awesomecua.com/resources/arxiv-2510-26098/", "external_url": "https://arxiv.org/abs/2510.26098", "title": "GUI Knowledge Bench: Revealing the Knowledge Gap of VLMs in GUI Tasks", "content_text": "GUI Knowledge Bench tests knowledge of widgets, interactions, and procedures across six named platforms and 292 applications to assess VLMs used for GUI tasks.", "summary": "GUI Knowledge Bench tests knowledge of widgets, interactions, and procedures across six named platforms and 292 applications to assess VLMs used for GUI tasks.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "evaluation", "screen-understanding", "state-tracking" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "web", "desktop", "windows", "macos", "linux", "android", "ios", "cross-platform" ], "record_json": "https://awesomecua.com/resources/arxiv-2510-26098/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2510-26098/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2509-23738/", "url": "https://awesomecua.com/resources/arxiv-2509-23738/", "external_url": "https://arxiv.org/abs/2509.23738", "title": "GUI-Shepherd: Reliable Process Reward and Verification for Long-Sequence GUI Tasks", "content_text": "Introduces GUI-Shepherd, a process reward model trained on annotated GUI interactions to give step-level feedback for agent training and verification on AndroidWorld and AndroidControl.", "summary": "Introduces GUI-Shepherd, a process reward model trained on annotated GUI interactions to give step-level feedback for agent training and verification on AndroidWorld and AndroidControl.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "reward-models", "reinforcement-learning", "evaluation" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "unspecified" ], "record_json": "https://awesomecua.com/resources/arxiv-2509-23738/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2509-23738/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2510-04791/", "url": "https://awesomecua.com/resources/arxiv-2510-04791/", "external_url": "https://arxiv.org/abs/2510.04791", "title": "GUISpector: An MLLM Agent Framework for Automated Verification of Natural Language Requirements in GUI Prototypes", "content_text": "Introduces GUISpector, an MLLM agent framework that operates GUI prototypes to check natural-language requirements and produces feedback for developers.", "summary": "Introduces GUISpector, an MLLM agent framework that operates GUI prototypes to check natural-language requirements and produces feedback for developers.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "automation", "planning", "observability" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "unspecified" ], "record_json": "https://awesomecua.com/resources/arxiv-2510-04791/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2510-04791/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2508-04010/", "url": "https://awesomecua.com/resources/arxiv-2508-04010/", "external_url": "https://arxiv.org/abs/2508.04010", "title": "HarmonyGuard: Toward Safety and Utility in Web Agents via Adaptive Policy Enhancement and Dual-Objective Optimization", "content_text": "Introduces HarmonyGuard, a multi-agent framework that updates security policies and weighs safety against task completion for agents operating on the open web.", "summary": "Introduces HarmonyGuard, a multi-agent framework that updates security policies and weighs safety against task completion for agents operating on the open web.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "safety-security", "web-navigation", "planning" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "web" ], "record_json": "https://awesomecua.com/resources/arxiv-2508-04010/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2508-04010/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2602-11514/", "url": "https://awesomecua.com/resources/arxiv-2602-11514/", "external_url": "https://arxiv.org/abs/2602.11514", "title": "How Smart Is Your GUI Agent? A Framework for the Future of Software Interaction", "content_text": "Proposes GUI Agent Autonomy Levels (GAL), a six-level framework for describing autonomy and benchmarking GUI agents that navigate web, desktop, and mobile software.", "summary": "Proposes GUI Agent Autonomy Levels (GAL), a six-level framework for describing autonomy and benchmarking GUI agents that navigate web, desktop, and mobile software.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "evaluation", "safety-security" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "web", "desktop" ], "record_json": "https://awesomecua.com/resources/arxiv-2602-11514/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2602-11514/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2602-10863/", "url": "https://awesomecua.com/resources/arxiv-2602-10863/", "external_url": "https://arxiv.org/abs/2602.10863", "title": "ICA: Information-Aware Credit Assignment for Visually Grounded Long-Horizon Information-Seeking Agents", "content_text": "Information-Aware Credit Assignment scores intermediate information gathered by web agents, using rendered webpage snapshots to compare evidence across trajectories.", "summary": "Information-Aware Credit Assignment scores intermediate information gathered by web agents, using rendered webpage snapshots to compare evidence across trajectories.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "reinforcement-learning", "reward-models", "web-navigation", "long-horizon" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "web" ], "record_json": "https://awesomecua.com/resources/arxiv-2602-10863/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2602-10863/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2509-07098/", "url": "https://awesomecua.com/resources/arxiv-2509-07098/", "external_url": "https://arxiv.org/abs/2509.07098", "title": "Instruction Agent: Enhancing Agent with Expert Demonstration", "content_text": "Introduces Instruction Agent, a GUI agent that follows instructions extracted from one expert demonstration and uses verification and backtracking to handle interruptions.", "summary": "Introduces Instruction Agent, a GUI agent that follows instructions extracted from one expert demonstration and uses verification and backtracking to handle interruptions.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "imitation-learning", "long-horizon", "state-tracking", "self-reflection" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "unspecified" ], "record_json": "https://awesomecua.com/resources/arxiv-2509-07098/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2509-07098/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2508-18265/", "url": "https://awesomecua.com/resources/arxiv-2508-18265/", "external_url": "https://arxiv.org/abs/2508.18265", "title": "InternVL3.5: Advancing Open-Source Multimodal Models in Versatility, Reasoning, and Efficiency", "content_text": "Introduces InternVL3.5, a multimodal model family trained with Cascade RL and designed for more efficient inference; GUI interaction is among its stated capabilities.", "summary": "Introduces InternVL3.5, a multimodal model family trained with Cascade RL and designed for more efficient inference; GUI interaction is among its stated capabilities.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "reinforcement-learning", "efficiency" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "unspecified" ], "record_json": "https://awesomecua.com/resources/arxiv-2508-18265/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2508-18265/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2510-03485/", "url": "https://awesomecua.com/resources/arxiv-2510-03485/", "external_url": "https://arxiv.org/abs/2510.03485", "title": "Learning Efficient Guardrails for Compliance", "content_text": "Introduces PolicyGuardBench for detecting policy violations in full web-agent trajectories and their prefixes, and trains PolicyGuard as an efficient guardrail model.", "summary": "Introduces PolicyGuardBench for detecting policy violations in full web-agent trajectories and their prefixes, and trains PolicyGuard as an efficient guardrail model.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "safety-security", "evaluation", "efficiency" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "web" ], "record_json": "https://awesomecua.com/resources/arxiv-2510-03485/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2510-03485/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2602-10525/", "url": "https://awesomecua.com/resources/arxiv-2602-10525/", "external_url": "https://arxiv.org/abs/2602.10525", "title": "LHAW: Controllable Underspecification for Long-Horizon Tasks", "content_text": "LHAW generates underspecified variants of long-horizon workflow tasks and uses agent trials to evaluate when missing information requires clarification.", "summary": "LHAW generates underspecified variants of long-horizon workflow tasks and uses agent trials to evaluate when missing information requires clarification.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "evaluation", "long-horizon", "human-in-the-loop", "data-generation" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "unspecified" ], "record_json": "https://awesomecua.com/resources/arxiv-2602-10525/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2602-10525/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2510-14969/", "url": "https://awesomecua.com/resources/arxiv-2510-14969/", "external_url": "https://arxiv.org/abs/2510.14969", "title": "LLMs as Scalable, General-Purpose Simulators For Evolving Digital Agent Training", "content_text": "Introduces UI-Simulator, which generates UI states and transitions for agent-training trajectories, and UI-Simulator-Grow, which prioritizes tasks and trajectory variants.", "summary": "Introduces UI-Simulator, which generates UI states and transitions for agent-training trajectories, and UI-Simulator-Grow, which prioritizes tasks and trajectory variants.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "world-models", "data-generation" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "unspecified" ], "record_json": "https://awesomecua.com/resources/arxiv-2510-14969/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2510-14969/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2601-18113/", "url": "https://awesomecua.com/resources/arxiv-2601-18113/", "external_url": "https://arxiv.org/abs/2601.18113", "title": "MalURLBench: A Benchmark Evaluating Agents' Vulnerabilities When Processing Web URLs", "content_text": "MalURLBench evaluates how LLM-based web agents handle disguised malicious URLs across attack scenarios; the paper also analyzes attack factors and proposes URLGuard.", "summary": "MalURLBench evaluates how LLM-based web agents handle disguised malicious URLs across attack scenarios; the paper also analyzes attack factors and proposes URLGuard.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "safety-security", "evaluation", "web-navigation" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "web" ], "record_json": "https://awesomecua.com/resources/arxiv-2601-18113/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2601-18113/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2509-17336/", "url": "https://awesomecua.com/resources/arxiv-2509-17336/", "external_url": "https://arxiv.org/abs/2509.17336", "title": "Mano Technical Report", "content_text": "Introduces Mano, a GUI agent trained with simulated data, supervised fine-tuning, and offline and online RL, with a verification module for error recovery.", "summary": "Introduces Mano, a GUI agent trained with simulated data, supervised fine-tuning, and offline and online RL, with a verification module for error recovery.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "data-generation", "reinforcement-learning", "self-reflection" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "unspecified" ], "record_json": "https://awesomecua.com/resources/arxiv-2509-17336/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2509-17336/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2511-23281/", "url": "https://awesomecua.com/resources/arxiv-2511-23281/", "external_url": "https://arxiv.org/abs/2511.23281", "title": "MCP vs RAG vs NLWeb vs HTML: A Comparison of the Effectiveness and Efficiency of Different Agent Interfaces to the Web (Technical Report)", "content_text": "Compares HTML browsing, RAG, MCP, and NLWeb interfaces for web agents performing product search, comparison, and checkout tasks in four simulated e-shops.", "summary": "Compares HTML browsing, RAG, MCP, and NLWeb interfaces for web agents performing product search, comparison, and checkout tasks in four simulated e-shops.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "evaluation", "tool-integration", "efficiency", "web-navigation" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "web" ], "record_json": "https://awesomecua.com/resources/arxiv-2511-23281/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2511-23281/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2601-20162/", "url": "https://awesomecua.com/resources/arxiv-2601-20162/", "external_url": "https://arxiv.org/abs/2601.20162", "title": "Me-Agent: A Personalized Mobile Agent with Two-Level User Habit Learning for Enhanced Interaction", "content_text": "Me-Agent personalizes mobile agent behavior through a preference reward model and hierarchical memory for long-term and app-specific habits, and introduces User FingerTip.", "summary": "Me-Agent personalizes mobile agent behavior through a preference reward model and hierarchical memory for long-term and app-specific habits, and introduces User FingerTip.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "mobile-navigation", "memory", "reward-models", "interaction-history" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "unspecified" ], "record_json": "https://awesomecua.com/resources/arxiv-2601-20162/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2601-20162/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2508-00935/", "url": "https://awesomecua.com/resources/arxiv-2508-00935/", "external_url": "https://arxiv.org/abs/2508.00935", "title": "Measuring Harmfulness of Computer-Using Agents", "content_text": "Introduces CUAHarm, 104 computer-use misuse tasks in a sandbox with verifiable rewards, and studies whether language models can monitor unsafe agent actions.", "summary": "Introduces CUAHarm, 104 computer-use misuse tasks in a sandbox with verifiable rewards, and studies whether language models can monitor unsafe agent actions.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "safety-security", "evaluation", "sandbox", "observability" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "unspecified" ], "record_json": "https://awesomecua.com/resources/arxiv-2508-00935/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2508-00935/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2510-20707/", "url": "https://awesomecua.com/resources/arxiv-2510-20707/", "external_url": "https://arxiv.org/abs/2510.20707", "title": "Mixing Importance with Diversity: Joint Optimization for KV Cache Compression in Large Vision-Language Models", "content_text": "MixKV compresses vision-language model KV caches by balancing importance and diversity; its evaluations include GUI grounding alongside broader multimodal tasks.", "summary": "MixKV compresses vision-language model KV caches by balancing importance and diversity; its evaluations include GUI grounding alongside broader multimodal tasks.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "efficiency", "grounding" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "unspecified" ], "record_json": "https://awesomecua.com/resources/arxiv-2510-20707/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2510-20707/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2508-13186/", "url": "https://awesomecua.com/resources/arxiv-2508-13186/", "external_url": "https://arxiv.org/abs/2508.13186", "title": "MM-BrowseComp: A Comprehensive Benchmark for Multimodal Browsing Agents", "content_text": "Introduces MM-BrowseComp, a benchmark for browsing agents that must use evidence from web images and videos to answer multimodal search questions.", "summary": "Introduces MM-BrowseComp, a benchmark for browsing agents that must use evidence from web images and videos to answer multimodal search questions.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "evaluation", "web-navigation", "video" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "web" ], "record_json": "https://awesomecua.com/resources/arxiv-2508-13186/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2508-13186/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2508-21475/", "url": "https://awesomecua.com/resources/arxiv-2508-21475/", "external_url": "https://arxiv.org/abs/2508.21475", "title": "MMSearch-Plus: Benchmarking Provenance-Aware Search for Multimodal Browsing Agents", "content_text": "Introduces MMSearch-Plus, a multimodal browsing benchmark with iterative image-text search, and an agent framework that marks and crops images for targeted retrieval.", "summary": "Introduces MMSearch-Plus, a multimodal browsing benchmark with iterative image-text search, and an agent framework that marks and crops images for targeted retrieval.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "evaluation", "web-navigation", "tool-integration", "long-horizon" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "web" ], "record_json": "https://awesomecua.com/resources/arxiv-2508-21475/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2508-21475/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2602-09222/", "url": "https://awesomecua.com/resources/arxiv-2602-09222/", "external_url": "https://arxiv.org/abs/2602.09222", "title": "MUZZLE: Adaptive Agentic Red-Teaming of Web Agents Against Indirect Prompt Injection Attacks", "content_text": "MUZZLE evaluates web-agent exposure to indirect prompt injection by finding attack surfaces in agent trajectories and refining injected instructions from execution feedback.", "summary": "MUZZLE evaluates web-agent exposure to indirect prompt injection by finding attack surfaces in agent trajectories and refining injected instructions from execution feedback.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "safety-security", "evaluation", "interaction-history", "web-navigation" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "web" ], "record_json": "https://awesomecua.com/resources/arxiv-2602-09222/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2602-09222/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2508-09124/", "url": "https://awesomecua.com/resources/arxiv-2508-09124/", "external_url": "https://arxiv.org/abs/2508.09124", "title": "OdysseyBench: Evaluating LLM Agents on Long-Horizon Complex Office Application Workflows", "content_text": "Introduces OdysseyBench for evaluating long-horizon workflows across office applications, plus HomerAgents to generate tasks from environment exploration and dialogue synthesis.", "summary": "Introduces OdysseyBench for evaluating long-horizon workflows across office applications, plus HomerAgents to generate tasks from environment exploration and dialogue synthesis.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "evaluation", "long-horizon", "cross-app", "interaction-history" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "unspecified" ], "record_json": "https://awesomecua.com/resources/arxiv-2508-09124/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2508-09124/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2601-22770/", "url": "https://awesomecua.com/resources/arxiv-2601-22770/", "external_url": "https://arxiv.org/abs/2601.22770", "title": "Okara: Detection and Attribution of TLS Man-in-the-Middle Vulnerabilities in Android Apps with Foundation Models", "content_text": "Okara uses foundation-model-driven GUI agents to explore Android apps for TLS vulnerability detection, then applies instrumentation and classification to attribute vulnerable code.", "summary": "Okara uses foundation-model-driven GUI agents to explore Android apps for TLS vulnerability detection, then applies instrumentation and classification to attribute vulnerable code.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "automation", "mobile-navigation", "safety-security" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "android" ], "record_json": "https://awesomecua.com/resources/arxiv-2601-22770/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2601-22770/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2601-02371/", "url": "https://awesomecua.com/resources/arxiv-2601-02371/", "external_url": "https://arxiv.org/abs/2601.02371", "title": "Permission Manifests for Web Agents", "content_text": "Proposes agent-permissions.json, a website manifest for declaring which interactions web agents may perform, with optional references to APIs.", "summary": "Proposes agent-permissions.json, a website manifest for declaring which interactions web agents may perform, with optional references to APIs.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "safety-security", "web-navigation", "tool-integration" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "web" ], "record_json": "https://awesomecua.com/resources/arxiv-2601-02371/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2601-02371/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2509-03536/", "url": "https://awesomecua.com/resources/arxiv-2509-03536/", "external_url": "https://arxiv.org/abs/2509.03536", "title": "PG-Agent: An Agent Powered by Page Graph", "content_text": "Introduces PG-Agent, which turns GUI operation episodes into page graphs and retrieves guidance from them to support task decomposition in new scenarios.", "summary": "Introduces PG-Agent, which turns GUI operation episodes into page graphs and retrieves guidance from them to support task decomposition in new scenarios.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "memory", "planning", "interaction-history", "screen-understanding" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "unspecified" ], "record_json": "https://awesomecua.com/resources/arxiv-2509-03536/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2509-03536/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2510-15863/", "url": "https://awesomecua.com/resources/arxiv-2510-15863/", "external_url": "https://arxiv.org/abs/2510.15863", "title": "PolySkill: Learning Generalizable Skills Through Polymorphic Abstraction", "content_text": "Introduces PolySkill, which separates a web-agent skill's goal from its execution so agents can learn and reuse skills across websites.", "summary": "Introduces PolySkill, which separates a web-agent skill's goal from its execution so agents can learn and reuse skills across websites.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "memory", "workflow", "web-navigation" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "web" ], "record_json": "https://awesomecua.com/resources/arxiv-2510-15863/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2510-15863/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2511-22074/", "url": "https://awesomecua.com/resources/arxiv-2511-22074/", "external_url": "https://arxiv.org/abs/2511.22074", "title": "Real-Time Procedural Learning From Experience for AI Agents", "content_text": "PRAXIS stores action outcomes and retrieves past state-action-result examples to guide agent decisions, with evaluation on the REAL web browsing benchmark.", "summary": "PRAXIS stores action outcomes and retrieves past state-action-result examples to guide agent decisions, with evaluation on the REAL web browsing benchmark.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "memory", "interaction-history", "web-navigation" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "web" ], "record_json": "https://awesomecua.com/resources/arxiv-2511-22074/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2511-22074/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2509-21072/", "url": "https://awesomecua.com/resources/arxiv-2509-21072/", "external_url": "https://arxiv.org/abs/2509.21072", "title": "Recon-Act: A Self-Evolving Multi-Agent Browser-Use System via Web Reconnaissance, Tool Generation, and Task Execution", "content_text": "Introduces Recon-Act, a browser-use framework that compares failed and successful web trajectories to create reusable tools for planning and executing long-horizon website tasks.", "summary": "Introduces Recon-Act, a browser-use framework that compares failed and successful web trajectories to create reusable tools for planning and executing long-horizon website tasks.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "web-navigation", "long-horizon", "planning", "tool-integration" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "web" ], "record_json": "https://awesomecua.com/resources/arxiv-2509-21072/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2509-21072/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2508-19461/", "url": "https://awesomecua.com/resources/arxiv-2508-19461/", "external_url": "https://arxiv.org/abs/2508.19461", "title": "Reliable Weak-to-Strong Monitoring of LLM Agents", "content_text": "Introduces CUA-SHADE-Arena and a red-teaming workflow for testing monitors that detect covert misbehavior in computer-use agent trajectories.", "summary": "Introduces CUA-SHADE-Arena and a red-teaming workflow for testing monitors that detect covert misbehavior in computer-use agent trajectories.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "safety-security", "evaluation", "observability" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "unspecified" ], "record_json": "https://awesomecua.com/resources/arxiv-2508-19461/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2508-19461/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2509-13313/", "url": "https://awesomecua.com/resources/arxiv-2509-13313/", "external_url": "https://arxiv.org/abs/2509.13313", "title": "ReSum: Unlocking Long-Horizon Search Intelligence via Context Summarization", "content_text": "Introduces ReSum, which compresses interaction histories during long web searches, and ReSum-GRPO, which trains agents to reason across the resulting segments.", "summary": "Introduces ReSum, which compresses interaction histories during long web searches, and ReSum-GRPO, which trains agents to reason across the resulting segments.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "web-navigation", "interaction-history", "efficiency", "reinforcement-learning" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "web" ], "record_json": "https://awesomecua.com/resources/arxiv-2509-13313/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2509-13313/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2510-14308/", "url": "https://awesomecua.com/resources/arxiv-2510-14308/", "external_url": "https://arxiv.org/abs/2510.14308", "title": "ReUseIt: Synthesizing Reusable AI Agent Workflows for Web Automation", "content_text": "Introduces ReUseIt, which synthesizes reusable web-agent workflows from successful and failed attempts and adds guards to detect and fix execution errors.", "summary": "Introduces ReUseIt, which synthesizes reusable web-agent workflows from successful and failed attempts and adds guards to detect and fix execution errors.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "workflow", "automation", "state-tracking" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "web" ], "record_json": "https://awesomecua.com/resources/arxiv-2510-14308/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2510-14308/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2510-02204/", "url": "https://awesomecua.com/resources/arxiv-2510-02204/", "external_url": "https://arxiv.org/abs/2510.02204", "title": "Say One Thing, Do Another? Diagnosing Reasoning-Execution Gaps in VLM-Powered Mobile-Use Agents", "content_text": "Introduces an evaluation framework for mobile GUI agents that compares actions implied by their reasoning with intended and executed actions to diagnose mismatches.", "summary": "Introduces an evaluation framework for mobile GUI agents that compares actions implied by their reasoning with intended and executed actions to diagnose mismatches.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "evaluation", "mobile-navigation", "observability", "safety-security" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "unspecified" ], "record_json": "https://awesomecua.com/resources/arxiv-2510-02204/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2510-02204/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2511-03773/", "url": "https://awesomecua.com/resources/arxiv-2511-03773/", "external_url": "https://arxiv.org/abs/2511.03773", "title": "Scaling Agent Learning via Experience Synthesis", "content_text": "DreamGym synthesizes agent experiences with a reasoning-based environment model for online RL; its evaluations include the web task setting WebArena.", "summary": "DreamGym synthesizes agent experiences with a reasoning-based environment model for online RL; its evaluations include the web task setting WebArena.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "data-generation", "reinforcement-learning", "world-models" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "unspecified" ], "record_json": "https://awesomecua.com/resources/arxiv-2511-03773/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2511-03773/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2507-16704/", "url": "https://awesomecua.com/resources/arxiv-2507-16704/", "external_url": "https://arxiv.org/abs/2507.16704", "title": "Screen2AX: Vision-Based Approach for Automatic macOS Accessibility Generation", "content_text": "Introduces Screen2AX, which creates hierarchical accessibility metadata from macOS screenshots, plus datasets and Screen2AX-Task for desktop-agent evaluation.", "summary": "Introduces Screen2AX, which creates hierarchical accessibility metadata from macOS screenshots, plus datasets and Screen2AX-Task for desktop-agent evaluation.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "screen-understanding", "data-generation", "evaluation" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "desktop", "macos" ], "record_json": "https://awesomecua.com/resources/arxiv-2507-16704/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2507-16704/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2510-19245/", "url": "https://awesomecua.com/resources/arxiv-2510-19245/", "external_url": "https://arxiv.org/abs/2510.19245", "title": "See, Think, Act: Online Shopper Behavior Simulation with VLM Agents", "content_text": "Studies online shopper action prediction with VLM agents using webpage screenshots, HTML, and action history from OPeRA, trained with SFT and hierarchical RL rewards.", "summary": "Studies online shopper action prediction with VLM agents using webpage screenshots, HTML, and action history from OPeRA, trained with SFT and hierarchical RL rewards.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "action-prediction", "interaction-history", "reinforcement-learning" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "web" ], "record_json": "https://awesomecua.com/resources/arxiv-2510-19245/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2510-19245/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2510-19247/", "url": "https://awesomecua.com/resources/arxiv-2510-19247/", "external_url": "https://arxiv.org/abs/2510.19247", "title": "SheetBrain: A Neuro-Symbolic Agent for Accurate Reasoning over Complex and Large Spreadsheets", "content_text": "Introduces SheetBrain, a spreadsheet question-answering and manipulation agent that uses a Python sandbox and Excel tools, plus the SheetBench benchmark.", "summary": "Introduces SheetBrain, a spreadsheet question-answering and manipulation agent that uses a Python sandbox and Excel tools, plus the SheetBench benchmark.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "tool-integration", "evaluation", "science" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "unspecified" ], "record_json": "https://awesomecua.com/resources/arxiv-2510-19247/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2510-19247/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2602-01443/", "url": "https://awesomecua.com/resources/arxiv-2602-01443/", "external_url": "https://arxiv.org/abs/2602.01443", "title": "SimGym: Traffic-Grounded Browser Agents for Offline A/B Testing in E-Commerce", "content_text": "SimGym runs traffic-grounded synthetic buyers as browser agents to compare e-commerce storefront variants offline, using buyer profiles derived from production interactions.", "summary": "SimGym runs traffic-grounded synthetic buyers as browser agents to compare e-commerce storefront variants offline, using buyer profiles derived from production interactions.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "ui-design", "web-navigation", "evaluation", "data-generation" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "web" ], "record_json": "https://awesomecua.com/resources/arxiv-2602-01443/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2602-01443/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2512-22322/", "url": "https://awesomecua.com/resources/arxiv-2512-22322/", "external_url": "https://arxiv.org/abs/2512.22322", "title": "SmartSnap: Proactive Evidence Seeking for Self-Verifying Agents", "content_text": "SmartSnap trains GUI agents to collect concise screenshot evidence of task completion for an LLM judge, with experiments on mobile tasks.", "summary": "SmartSnap trains GUI agents to collect concise screenshot evidence of task completion for an LLM judge, with experiments on mobile tasks.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "self-reflection", "evaluation", "reinforcement-learning", "mobile-navigation" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "unspecified" ], "record_json": "https://awesomecua.com/resources/arxiv-2512-22322/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2512-22322/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2601-22491/", "url": "https://awesomecua.com/resources/arxiv-2601-22491/", "external_url": "https://arxiv.org/abs/2601.22491", "title": "SSL: Sweet Spot Learning for Differentiated Guidance in Agentic Optimization", "content_text": "SSL uses progressively tiered rewards for agent training, including distance-based guidance on GUI perception tasks alongside planning and reasoning tasks.", "summary": "SSL uses progressively tiered rewards for agent training, including distance-based guidance on GUI perception tasks alongside planning and reasoning tasks.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "reinforcement-learning", "reward-models", "efficiency" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "unspecified" ], "record_json": "https://awesomecua.com/resources/arxiv-2601-22491/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2601-22491/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2510-14513/", "url": "https://awesomecua.com/resources/arxiv-2510-14513/", "external_url": "https://arxiv.org/abs/2510.14513", "title": "State Your Intention to Steer Your Attention: An AI Assistant for Intentional Digital Living", "content_text": "Introduces an assistant that reads screenshots, application titles, and URLs to compare computer activity with a user's stated intention and send notifications when it diverges.", "summary": "Introduces an assistant that reads screenshots, application titles, and URLs to compare computer activity with a user's stated intention and send notifications when it diverges.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "screen-understanding", "human-in-the-loop", "state-tracking" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "unspecified" ], "record_json": "https://awesomecua.com/resources/arxiv-2510-14513/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2510-14513/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2511-13091/", "url": "https://awesomecua.com/resources/arxiv-2511-13091/", "external_url": "https://arxiv.org/abs/2511.13091", "title": "STEP: Success-Rate-Aware Trajectory-Efficient Policy Optimization", "content_text": "STEP allocates RL sampling by task success rate and trains on individual steps from interaction trajectories, with evaluations on OSWorld and AndroidWorld.", "summary": "STEP allocates RL sampling by task success rate and trains on individual steps from interaction trajectories, with evaluations on OSWorld and AndroidWorld.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "reinforcement-learning", "efficiency", "long-horizon" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "unspecified" ], "record_json": "https://awesomecua.com/resources/arxiv-2511-13091/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2511-13091/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2508-16271/", "url": "https://awesomecua.com/resources/arxiv-2508-16271/", "external_url": "https://arxiv.org/abs/2508.16271", "title": "Structuring GUI Elements through Vision Language Models: Towards Action Space Generation", "content_text": "Introduces IAML training, which augments coordinate samples to help multimodal models locate GUI elements from screen content and user instructions.", "summary": "Introduces IAML training, which augments coordinate samples to help multimodal models locate GUI elements from screen content and user instructions.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "grounding", "screen-understanding", "data-generation" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "unspecified" ], "record_json": "https://awesomecua.com/resources/arxiv-2508-16271/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2508-16271/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2508-20018/", "url": "https://awesomecua.com/resources/arxiv-2508-20018/", "external_url": "https://arxiv.org/abs/2508.20018", "title": "SWIRL: A Staged Workflow for Interleaved Reinforcement Learning in Mobile GUI Control", "content_text": "Introduces SWIRL, staged interleaved RL for mobile GUI control, pairing a Navigator that plans from screens and language with an Interactor that produces actions.", "summary": "Introduces SWIRL, staged interleaved RL for mobile GUI control, pairing a Navigator that plans from screens and language with an Interactor that produces actions.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "reinforcement-learning", "planning", "grounding", "mobile-navigation" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "unspecified" ], "record_json": "https://awesomecua.com/resources/arxiv-2508-20018/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2508-20018/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2511-09309/", "url": "https://awesomecua.com/resources/arxiv-2511-09309/", "external_url": "https://arxiv.org/abs/2511.09309", "title": "TaskSense: Cognitive Chain Modeling and Difficulty Estimation for GUI Tasks", "content_text": "Cognitive Chain estimates GUI task difficulty by extracting sequences of cognitive steps from execution traces, then uses those estimates to assess GUI agents.", "summary": "Cognitive Chain estimates GUI task difficulty by extracting sequences of cognitive steps from execution traces, then uses those estimates to assess GUI agents.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "evaluation", "observability", "data-generation" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "unspecified" ], "record_json": "https://awesomecua.com/resources/arxiv-2511-09309/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2511-09309/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2601-11868/", "url": "https://awesomecua.com/resources/arxiv-2601-11868/", "external_url": "https://arxiv.org/abs/2601.11868", "title": "Terminal-Bench: Benchmarking Agents on Hard, Realistic Tasks in Command Line Interfaces", "content_text": "Terminal-Bench 2.0 benchmarks agents on 89 workflow-inspired tasks in computer terminal environments, with a distinct environment, human solution and verification tests for each task.", "summary": "Terminal-Bench 2.0 benchmarks agents on 89 workflow-inspired tasks in computer terminal environments, with a distinct environment, human solution and verification tests for each task.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "evaluation", "long-horizon", "workflow" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "unspecified" ], "record_json": "https://awesomecua.com/resources/arxiv-2601-11868/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2601-11868/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2509-21733/", "url": "https://awesomecua.com/resources/arxiv-2509-21733/", "external_url": "https://arxiv.org/abs/2509.21733", "title": "UISim: An Interactive Image-Based UI Simulator for Dynamic Mobile Environments", "content_text": "Introduces UISim, an image-based mobile UI simulator that predicts the next screen layout after an action and renders the resulting screen for testing and data generation.", "summary": "Introduces UISim, an image-based mobile UI simulator that predicts the next screen layout after an action and renders the resulting screen for testing and data generation.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "world-models", "sandbox", "data-generation", "mobile-navigation" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "unspecified" ], "record_json": "https://awesomecua.com/resources/arxiv-2509-21733/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2509-21733/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2510-04452/", "url": "https://awesomecua.com/resources/arxiv-2510-04452/", "external_url": "https://arxiv.org/abs/2510.04452", "title": "Understanding User Experiences of Computer Use Agents: Design Space and Opportunities for Building Agent UX Prototypes", "content_text": "Maps design considerations for computer-use agent experiences and introduces AgentUXlab, a browser-based probe for prototyping and evaluating agent UX on a website.", "summary": "Maps design considerations for computer-use agent experiences and introduces AgentUXlab, a browser-based probe for prototyping and evaluating agent UX on a website.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "ui-design", "human-in-the-loop", "evaluation" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "web" ], "record_json": "https://awesomecua.com/resources/arxiv-2510-04452/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2510-04452/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2508-04026/", "url": "https://awesomecua.com/resources/arxiv-2508-04026/", "external_url": "https://arxiv.org/abs/2508.04026", "title": "VeriWeb: Verifiable Long-Chain Web Benchmark for Agentic Information-Seeking", "content_text": "Introduces VeriWeb, a benchmark of long-chain web information-seeking tasks with verifiable subtasks and a complete human-annotated trajectory for each task.", "summary": "Introduces VeriWeb, a benchmark of long-chain web information-seeking tasks with verifiable subtasks and a complete human-annotated trajectory for each task.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "evaluation", "long-horizon", "data-generation" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "web" ], "record_json": "https://awesomecua.com/resources/arxiv-2508-04026/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2508-04026/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2510-04673/", "url": "https://awesomecua.com/resources/arxiv-2510-04673/", "external_url": "https://arxiv.org/abs/2510.04673", "title": "Watch and Learn: Learning to Use Computers from Online Videos", "content_text": "Introduces Watch & Learn (W&L), which converts human computer-use videos into UI action trajectories by predicting actions between consecutive screen states.", "summary": "Introduces Watch & Learn (W&L), which converts human computer-use videos into UI action trajectories by predicting actions between consecutive screen states.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "video", "data-generation", "action-prediction" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "unspecified" ], "record_json": "https://awesomecua.com/resources/arxiv-2510-04673/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2510-04673/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2510-14438/", "url": "https://awesomecua.com/resources/arxiv-2510-14438/", "external_url": "https://arxiv.org/abs/2510.14438", "title": "WebAggregator: Enhancing Compositional Reasoning Capabilities of Deep Research Agent Foundation Models", "content_text": "Introduces WebAggregator, a pipeline for synthesizing compositional questions from websites to train deep-research agents, plus the WebAggregatorQA testbed.", "summary": "Introduces WebAggregator, a pipeline for synthesizing compositional questions from websites to train deep-research agents, plus the WebAggregatorQA testbed.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "data-generation", "evaluation", "web-navigation" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "web" ], "record_json": "https://awesomecua.com/resources/arxiv-2510-14438/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2510-14438/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2510-22732/", "url": "https://awesomecua.com/resources/arxiv-2510-22732/", "external_url": "https://arxiv.org/abs/2510.22732", "title": "WebATLAS: An LLM Agent with Experience-Driven Memory and Action Simulation", "content_text": "WebATLAS is a web agent that stores interaction outcomes and simulates candidate actions before acting, using past experience to plan navigation on unseen websites.", "summary": "WebATLAS is a web agent that stores interaction outcomes and simulates candidate actions before acting, using past experience to plan navigation on unseen websites.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "memory", "world-models", "planning", "web-navigation" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "web" ], "record_json": "https://awesomecua.com/resources/arxiv-2510-22732/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2510-22732/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2509-22644/", "url": "https://awesomecua.com/resources/arxiv-2509-22644/", "external_url": "https://arxiv.org/abs/2509.22644", "title": "WebGen-Agent: Enhancing Interactive Website Generation with Multi-Level Feedback and Step-Level Reinforcement Learning", "content_text": "Introduces WebGen-Agent for website code generation, using screenshot critiques and GUI-agent tests to refine code and provide step-level training rewards.", "summary": "Introduces WebGen-Agent for website code generation, using screenshot critiques and GUI-agent tests to refine code and provide step-level training rewards.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "ui-design", "reward-models", "reinforcement-learning", "self-reflection" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "web" ], "record_json": "https://awesomecua.com/resources/arxiv-2509-22644/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2509-22644/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2508-16763/", "url": "https://awesomecua.com/resources/arxiv-2508-16763/", "external_url": "https://arxiv.org/abs/2508.16763", "title": "WebMMU: A Benchmark for Multimodal Multilingual Website Understanding and Code Generation", "content_text": "Introduces WebMMU, a multilingual benchmark for website visual question answering, element grounding, code editing, and mockup-to-code generation.", "summary": "Introduces WebMMU, a multilingual benchmark for website visual question answering, element grounding, code editing, and mockup-to-code generation.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "evaluation", "grounding", "screen-understanding", "ui-design" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "web" ], "record_json": "https://awesomecua.com/resources/arxiv-2508-16763/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2508-16763/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2512-12692/", "url": "https://awesomecua.com/resources/arxiv-2512-12692/", "external_url": "https://arxiv.org/abs/2512.12692", "title": "WebOperator: Action-Aware Tree Search for Autonomous Agents in Web Environment", "content_text": "WebOperator uses tree search, action filtering, and path checks before replay to explore web tasks and backtrack while accounting for irreversible actions.", "summary": "WebOperator uses tree search, action filtering, and path checks before replay to explore web tasks and backtrack while accounting for irreversible actions.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "planning", "safety-security", "web-navigation" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "web" ], "record_json": "https://awesomecua.com/resources/arxiv-2512-12692/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2512-12692/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2510-11221/", "url": "https://awesomecua.com/resources/arxiv-2510-11221/", "external_url": "https://arxiv.org/abs/2510.11221", "title": "WebRouter: Query-specific Router via Variational Information Bottleneck for Cost-sensitive Web Agent", "content_text": "Introduces WebRouter, a cost-aware router that compresses web-agent prompts to choose among LLMs, with cost and accuracy measured on WebVoyager websites.", "summary": "Introduces WebRouter, a cost-aware router that compresses web-agent prompts to choose among LLMs, with cost and accuracy measured on WebVoyager websites.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "efficiency", "web-navigation" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "web" ], "record_json": "https://awesomecua.com/resources/arxiv-2510-11221/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2510-11221/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2511-06251/", "url": "https://awesomecua.com/resources/arxiv-2511-06251/", "external_url": "https://arxiv.org/abs/2511.06251", "title": "WebVIA: A Web-based Vision-Language Agentic Framework for Interactive and Verifiable UI-to-Code Generation", "content_text": "WebVIA combines a web UI exploration agent, interactive UI-to-Code generation, and validation of generated code's interactivity.", "summary": "WebVIA combines a web UI exploration agent, interactive UI-to-Code generation, and validation of generated code's interactivity.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "ui-design", "evaluation", "web-navigation" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "web" ], "record_json": "https://awesomecua.com/resources/arxiv-2511-06251/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2511-06251/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2602-08235/", "url": "https://awesomecua.com/resources/arxiv-2602-08235/", "external_url": "https://arxiv.org/abs/2602.08235", "title": "When Benign Inputs Lead to Severe Harms: Eliciting Unsafe Unintended Behaviors of Computer-Use Agents", "content_text": "AutoElicit iteratively revises benign instructions using execution feedback to surface harmful unintended actions by computer-use agents in OS workflows.", "summary": "AutoElicit iteratively revises benign instructions using execution feedback to surface harmful unintended actions by computer-use agents in OS workflows.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "safety-security", "evaluation", "data-generation", "workflow" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "unspecified" ], "record_json": "https://awesomecua.com/resources/arxiv-2602-08235/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2602-08235/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2602-07008/", "url": "https://awesomecua.com/resources/arxiv-2602-07008/", "external_url": "https://arxiv.org/abs/2602.07008", "title": "Where Not to Learn: Prior-Aligned Training with Subset-based Attribution Constraints", "content_text": "Proposes training with subset-based attribution constraints so models attend to intended evidence, evaluated on image classification and GUI-agent click decisions.", "summary": "Proposes training with subset-based attribution constraints so models attend to intended evidence, evaluated on image classification and GUI-agent click decisions.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "action-prediction", "screen-understanding" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "unspecified" ], "record_json": "https://awesomecua.com/resources/arxiv-2602-07008/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2602-07008/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2602-15384/", "url": "https://awesomecua.com/resources/arxiv-2602-15384/", "external_url": "https://arxiv.org/abs/2602.15384", "title": "World-Model-Augmented Web Agents with Action Correction", "content_text": "WAC has a web-agent action model consult a world model, simulate action outcomes and use judge feedback to correct actions before execution.", "summary": "WAC has a web-agent action model consult a world model, simulate action outcomes and use judge feedback to correct actions before execution.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "world-models", "planning", "self-reflection", "web-navigation" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "web" ], "record_json": "https://awesomecua.com/resources/arxiv-2602-15384/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2602-15384/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2510-06607/", "url": "https://awesomecua.com/resources/arxiv-2510-06607/", "external_url": "https://arxiv.org/abs/2510.06607", "title": "Your Harness is Not Secure: Benchmarking Real-world Threat of Command Line Interface Agent", "content_text": "Introduces AdvCLI, a sandbox benchmark for evaluating whether command-line agents carry out security-relevant operating-system tasks across multiple hosts.", "summary": "Introduces AdvCLI, a sandbox benchmark for evaluating whether command-line agents carry out security-relevant operating-system tasks across multiple hosts.", "date_published": "2026-11-06T08:25:50Z", "date_modified": "2026-11-06T08:25:50Z", "tags": [ "paper", "evaluation", "safety-security", "sandbox" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "unspecified" ], "record_json": "https://awesomecua.com/resources/arxiv-2510-06607/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2510-06607/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-1812-11470/", "url": "https://awesomecua.com/resources/arxiv-1812-11470/", "external_url": "https://arxiv.org/abs/1812.11470", "title": "A Systematic Literature Review of Automated Techniques for Functional GUI Testing of Mobile Applications", "content_text": "Surveys automated functional GUI testing techniques for mobile apps, comparing the effectiveness, efficiency and practicality of approaches reported in 25 studies.", "summary": "Surveys automated functional GUI testing techniques for mobile apps, comparing the effectiveness, efficiency and practicality of approaches reported in 25 studies.", "date_published": "2026-11-06T07:33:19Z", "date_modified": "2026-11-06T07:33:19Z", "tags": [ "survey", "survey", "automation", "evaluation" ], "_awesomecua": { "kind": "survey", "review_status": "curated", "environments": [ "unspecified" ], "record_json": "https://awesomecua.com/resources/arxiv-1812-11470/record.json", "markdown": "https://awesomecua.com/resources/arxiv-1812-11470/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2507-10202/", "url": "https://awesomecua.com/resources/arxiv-2507-10202/", "external_url": "https://arxiv.org/abs/2507.10202", "title": "A Training-Free, Task-Agnostic Framework for Enhancing MLLM Performance on High-Resolution Images", "content_text": "Introduces Extract Candidate then Predict (ECP), a training-free two-stage method for high-resolution images, evaluated partly on 4K GUI grounding.", "summary": "Introduces Extract Candidate then Predict (ECP), a training-free two-stage method for high-resolution images, evaluated partly on 4K GUI grounding.", "date_published": "2026-11-06T07:33:19Z", "date_modified": "2026-11-06T07:33:19Z", "tags": [ "paper", "training-free", "grounding", "screen-understanding" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "unspecified" ], "record_json": "https://awesomecua.com/resources/arxiv-2507-10202/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2507-10202/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2502-14496/", "url": "https://awesomecua.com/resources/arxiv-2502-14496/", "external_url": "https://arxiv.org/abs/2502.14496", "title": "Advancing Language Multi-Agent Learning with Credit Re-Assignment for Interactive Environment Generalization", "content_text": "Introduces CollabUIAgents, a multi-agent RL framework that assigns LLM-based process rewards to help agents generalize across interactive settings including mobile operations and web browsing.", "summary": "Introduces CollabUIAgents, a multi-agent RL framework that assigns LLM-based process rewards to help agents generalize across interactive settings including mobile operations and web browsing.", "date_published": "2026-11-06T07:33:19Z", "date_modified": "2026-11-06T07:33:19Z", "tags": [ "paper", "reinforcement-learning", "reward-models", "robustness" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "web" ], "record_json": "https://awesomecua.com/resources/arxiv-2502-14496/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2502-14496/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2408-07199/", "url": "https://awesomecua.com/resources/arxiv-2408-07199/", "external_url": "https://arxiv.org/abs/2408.07199", "title": "Agent Q: Advanced Reasoning and Learning for Autonomous AI Agents", "content_text": "Agent Q combines guided tree search, self-critique and preference-based fine-tuning to train web agents from successful and unsuccessful interactions.", "summary": "Agent Q combines guided tree search, self-critique and preference-based fine-tuning to train web agents from successful and unsuccessful interactions.", "date_published": "2026-11-06T07:33:19Z", "date_modified": "2026-11-06T07:33:19Z", "tags": [ "paper", "planning", "self-reflection", "imitation-learning", "web-navigation" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "web" ], "record_json": "https://awesomecua.com/resources/arxiv-2408-07199/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2408-07199/index.md" } }, { "id": "https://awesomecua.com/resources/arxiv-2506-21252/", "url": "https://awesomecua.com/resources/arxiv-2506-21252/", "external_url": "https://arxiv.org/abs/2506.21252", "title": "Agent-RewardBench: Towards a Unified Benchmark for Reward Modeling across Perception, Planning, and Safety in Real-World Multimodal Agents", "content_text": "Introduces Agent-RewardBench to evaluate step-level reward models for multimodal agents across perception, planning, and safety; web navigation is among its scenarios.", "summary": "Introduces Agent-RewardBench to evaluate step-level reward models for multimodal agents across perception, planning, and safety; web navigation is among its scenarios.", "date_published": "2026-11-06T07:33:19Z", "date_modified": "2026-11-06T07:33:19Z", "tags": [ "paper", "evaluation", "reward-models", "planning", "safety-security" ], "_awesomecua": { "kind": "paper", "review_status": "curated", "environments": [ "web" ], "record_json": "https://awesomecua.com/resources/arxiv-2506-21252/record.json", "markdown": "https://awesomecua.com/resources/arxiv-2506-21252/index.md" } } ] }