Cover Image

Say One Thing, Do Another? Diagnosing Reasoning-Execution Gaps in VLM-Powered Mobile-Use Agents

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Dong, Lingzhong, Zhou, Ziqi, Yang, Shuaibo, Sheng, Haiyue, Cheng, Pengzhou, Wu, Zongru, Wu, Zheng, Liu, Gongshen, Zhang, Zhuosheng
Format:	Preprint
Published:	2025
Subjects:	Computation and Language
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

Similar Items

Agent-ScanKit: Unraveling Memory and Reasoning of Multimodal Agents via Sensitivity Perturbations
by: Cheng, Pengzhou, et al.
Published: (2025)

GEM: Gaussian Embedding Modeling for Out-of-Distribution Detection in GUI Agents
by: Wu, Zheng, et al.
Published: (2025)

Smoothing Grounding and Reasoning for MLLM-Powered GUI Agents with Query-Oriented Pivot Tasks
by: Wu, Zongru, et al.
Published: (2025)

Faithful Mobile GUI Agents with Guided Advantage Estimator
by: Hu, Haowen, et al.
Published: (2026)

See, Think, Act: Teaching Multimodal Agents to Effectively Interact with GUI by Identifying Toggles
by: Wu, Zongru, et al.
Published: (2025)

Hidden Ghost Hand: Unveiling Backdoor Vulnerabilities in MLLM-Powered Mobile GUI Agents
by: Cheng, Pengzhou, et al.
Published: (2025)

OS-Kairos: Adaptive Interaction for MLLM-Powered GUI Agents
by: Cheng, Pengzhou, et al.
Published: (2025)

Acquiring Clean Language Models from Backdoor Poisoned Datasets by Downscaling Frequency Space
by: Wu, Zongru, et al.
Published: (2024)

Gracefully Filtering Backdoor Samples for Generative Large Language Models without Retraining
by: Wu, Zongru, et al.
Published: (2024)

MKF-ADS: Multi-Knowledge Fusion Based Self-supervised Anomaly Detection System for Control Area Network
by: Cheng, Pengzhou, et al.
Published: (2024)

OS-SPEAR: A Toolkit for the Safety, Performance,Efficiency, and Robustness Analysis of OS Agents
by: Wu, Zheng, et al.
Published: (2026)

Transferring Backdoors between Large Language Models by Knowledge Distillation
by: Cheng, Pengzhou, et al.
Published: (2024)

SynGhost: Invisible and Universal Task-agnostic Backdoor Attack via Syntactic Transfer
by: Cheng, Pengzhou, et al.
Published: (2024)

TrojanRAG: Retrieval-Augmented Generation Can Be Backdoor Driver in Large Language Models
by: Cheng, Pengzhou, et al.
Published: (2024)

Large Language Models Often Say One Thing and Do Another
by: Xu, Ruoxi, et al.
Published: (2025)

When Agents Say One Thing and Do Another: Validating Elicited Beliefs from LLMs
by: Yamin, Khurram, et al.
Published: (2026)

Backdoor Attacks and Countermeasures in Natural Language Processing Models: A Comprehensive Security Review
by: Cheng, Pengzhou, et al.
Published: (2023)

On the Adaptive Psychological Persuasion of Large Language Models
by: Ju, Tianjie, et al.
Published: (2025)

Atomic-to-Compositional Generalization for Mobile Agents with A New Benchmark and Scheduling System
by: Guo, Yuan, et al.
Published: (2025)

When Disagreements Elicit Robustness: Investigating Self-Repair Capabilities under LLM Multi-Agent Disagreements
by: Ju, Tianjie, et al.
Published: (2025)

Training High-Level Schedulers with Execution-Feedback Reinforcement Learning for Long-Horizon GUI Automation
by: Deng, Zehao, et al.
Published: (2025)

VeriOS: Query-Driven Proactive Human-Agent-GUI Interaction for Trustworthy OS Agents
by: Wu, Zheng, et al.
Published: (2025)

One Thing Follows Another
by: Rosenthal, Sarah
Published: (2025)

Disagreements in Reasoning: How a Model's Thinking Process Dictates Persuasion in Multi-Agent Systems
by: Zhao, Haodong, et al.
Published: (2025)

Doing What They Say, Not What They Reason: Locating the Faithfulness Gap in LLM Agents
by: Wang, Yufeng
Published: (2026)

LaSM: Layer-wise Scaling Mechanism for Defending Pop-up Attack on GUI Agents
by: Yan, Zihe, et al.
Published: (2025)

ProtegoFed: Backdoor-Free Federated Instruction Tuning with Interspersed Poisoned Data
by: Zhao, Haodong, et al.
Published: (2026)

MindPower: Enabling Theory-of-Mind Reasoning in VLM-based Embodied Agents
by: Zhang, Ruoxuan, et al.
Published: (2025)

Flooding Spread of Manipulated Knowledge in LLM-Based Multi-Agent Communities
by: Ju, Tianjie, et al.
Published: (2024)

R-Judge: Benchmarking Safety Risk Awareness for LLM Agents
by: Yuan, Tongxin, et al.
Published: (2024)

Thinking in a Crowd: How Auxiliary Information Shapes LLM Reasoning
by: Zhao, Haodong, et al.
Published: (2025)

Say the Mission, Execute the Swarm: Agent-Enhanced LLM Reasoning in the Web-of-Drones
by: Iannoli, Andrea, et al.
Published: (2026)

One Thing Leads to Another: Faculty Outreach through Internet Instruction.
by: Borowske, Kate, et al.
Published: (1997)

And Another Thing: Nonanthropocentrism and Art
by: Behar, Katherine, et al.
Published: (2026)

And Another Thing: Nonanthropocentrism and Art
Published: (2019)

Quick on the Uptake: Eliciting Implicit Intents from Human Demonstrations for Personalized Mobile-Use Agents
by: Wu, Zheng, et al.
Published: (2025)

Mitigating Misleading Chain-of-Thought Reasoning with Selective Filtering
by: Wu, Yexin, et al.
Published: (2024)

Revealing the Gap in Human and VLM Scene Perception through Counterfactual Semantic Saliency
by: Wen, Ziqi, et al.
Published: (2026)

Spectrum of the Dirac operator on Compact Riemannian Manifolds
by: Zeng, Lingzhong
Published: (2024)

The First Eigenvalue of Embedded Minimal Hypersurfaces in the Unit Sphere I: Yau's Conjecture
by: Zeng, Lingzhong
Published: (2025)