arXiv subjects

Mengyan Wang

Publications and source records attributed to Mengyan Wang.

6 recordsLinked to original sources

arXiv2026

MobileForge: Annotation-Free Adaptation for Mobile GUI Agents with Hierarchical Feedback-Guided Policy Optimization

MLLM-based mobile GUI agents have made substantial progress in UI understanding and action execution, but adapting them to real target apps remains costly because mobile apps are numerous, frequently updated, and hard to cover with human-written tasks, demonstrations, or reward labels. Existing annotation-free GUI learning reduces manual supervision, yet lacks a unified substrate connecting target-app exploration, curriculum mining, rollout execution, and feedback, while policy optimization often relies on isolated rollouts and coarse rewards that are hard to convert into reliable improvement signals. We present MobileForge, an annotation-free adaptation system for mobile GUI agents. MobileForge consists of MobileGym, which grounds task generation and rollout evaluation in real mobile app interaction, and Hierarchical Feedback-Guided Policy Optimization (HiFPO), which turns trajectory outcomes, step-level process feedback, and corrective hints into hint-contextualized step-level GRPO updates. Using only automatically generated annotation-free adaptation data, MobileForge adapts Qwen3-VL-8B to 67.2% Pass@3 on AndroidWorld, close to the closed-data GUI-specialized GUI-Owl-1.5-8B base model at 69.0%. The MobileForge-adapted ForgeOwl-8B further reaches 77.6% Pass@3 on AndroidWorld and 41.0% success on the out-of-domain MobileWorld GUI-only split, establishing the strongest open-data mobile GUI agent in our evaluation. Code, data, and trained models will be released at https://mobile-forge.github.io/.

cs.HC↗

arXiv2026

MemGUI-Agent: An End-to-End Long-Horizon Mobile GUI Agent with Proactive Context Management

MLLM-based mobile GUI agents have made substantial progress on short-horizon tasks, yet remain unreliable on long-horizon tasks that require retaining intermediate facts across many steps and app transitions. We attribute this limitation to ReAct-style prompting, which passively accumulates per-step records, leading to prompt explosion and dilution of critical cross-app facts. To address this, we introduce MemGUI-Agent, an end-to-end long-horizon mobile GUI agent with proactive context management. MemGUI-Agent is built on Context-as-Action (ConAct), which casts context management as first-class actions emitted by the same policy that selects UI actions. Instead of passively appending history, ConAct maintains three structured context fields: folded action history, folded UI state, and recent step record, preserving critical UI facts while keeping context compact. To make proactive context management learnable across model scales, we construct MemGUI-3K, a 2,956-trajectory dataset with full ConAct annotations for supervised training and offline analysis. Training an 8B model on MemGUI-3K produces MemGUI-8B-SFT, an 8B MemGUI-Agent that achieves the best open-data 8B performance on MemGUI-Bench and generalizes to the out-of-distribution MobileWorld benchmark. Code, data, and trained models will be released at https://memgui-agent.github.io/.

cs.HC↗

arXiv2026

From Human Cognition to Neural Activations: Probing the Computational Primitives of Spatial Reasoning in LLMs

As spatial intelligence becomes an increasingly important capability for foundation models, it remains unclear whether large language models' (LLMs) performance on spatial reasoning benchmarks reflects structured internal spatial representations or reliance on linguistic heuristics. We address this question from a mechanistic perspective by examining how spatial information is internally represented and used. Drawing on computational theories of human spatial cognition, we decompose spatial reasoning into three primitives, relational composition, representational transformation, and stateful spatial updating, and design controlled task families for each. We evaluate multilingual LLMs in English, Chinese, and Arabic under single pass inference, and analyze internal representations using linear probing, sparse autoencoder based feature analysis, and causal interventions. We find that task relevant spatial information is encoded in intermediate layers and can causally influence behavior, but these representations are transient, fragmented across task families, and weakly integrated into final predictions. Cross linguistic analysis further reveals mechanistic degeneracy, where similar behavioral performance arises from distinct internal pathways. Overall, our results suggest that current LLMs exhibit limited and context dependent spatial representations rather than robust, general purpose spatial reasoning, highlighting the need for mechanistic evaluation beyond benchmark accuracy.

cs.CL↗

arXiv2024

Balancing Information Perception with Yin-Yang: Agent-Based Information Neutrality Model for Recommendation Systems

While preference-based recommendation algorithms effectively enhance user engagement by recommending personalized content, they often result in the creation of ``filter bubbles''. These bubbles restrict the range of information users interact with, inadvertently reinforcing their existing viewpoints. Previous research has focused on modifying these underlying algorithms to tackle this issue. Yet, approaches that maintain the integrity of the original algorithms remain largely unexplored. This paper introduces an Agent-based Information Neutrality model grounded in the Yin-Yang theory, namely, AbIN. This innovative approach targets the imbalance in information perception within existing recommendation systems. It is designed to integrate with these preference-based systems, ensuring the delivery of recommendations with neutral information. Our empirical evaluation of this model proved its efficacy, showcasing its capacity to expand information diversity while respecting user preferences. Consequently, AbIN emerges as an instrumental tool in mitigating the negative impact of filter bubbles on information consumption.

cs.HC↗

arXiv2023

BHEISR: Nudging from Bias to Balance -- Promoting Belief Harmony by Eliminating Ideological Segregation in Knowledge-based Recommendations

In the realm of personalized recommendation systems, the increasing concern is the amplification of belief imbalance and user biases, a phenomenon primarily attributed to the filter bubble. Addressing this critical issue, we introduce an innovative intermediate agency (BHEISR) between users and existing recommendation systems to attenuate the negative repercussions of the filter bubble effect in extant recommendation systems. The main objective is to strike a belief balance for users while minimizing the detrimental influence caused by filter bubbles. The BHEISR model amalgamates principles from nudge theory while upholding democratic and transparent principles. It harnesses user-specific category information to stimulate curiosity, even in areas users might initially deem uninteresting. By progressively stimulating interest in novel categories, the model encourages users to broaden their belief horizons and explore the information they typically overlook. Our model is time-sensitive and operates on a user feedback loop. It utilizes the existing recommendation algorithm of the model and incorporates user feedback from the prior time frame. This approach endeavors to transcend the constraints of the filter bubble, enrich recommendation diversity, and strike a belief balance among users while also catering to user preferences and system-specific business requirements. To validate the effectiveness and reliability of the BHEISR model, we conducted a series of comprehensive experiments with real-world datasets. These experiments compared the performance of the BHEISR model against several baseline models using nearly 200 filter bubble-impacted users as test subjects. Our experimental results conclusively illustrate the superior performance of the BHEISR model in mitigating filter bubbles and balancing user perspectives.

cs.IR↗

arXiv2023

DOR: A Novel Dual-Observation-Based Approach for News Recommendation Systems

Online social media platforms offer access to a vast amount of information, but sifting through the abundance of news can be overwhelming and tiring for readers. personalised recommendation algorithms can help users find information that interests them. However, most existing models rely solely on observations of user behaviour, such as viewing history, ignoring the connections between the news and a user's prior knowledge. This can result in a lack of diverse recommendations for individuals. In this paper, we propose a novel method to address the complex problem of news recommendation. Our approach is based on the idea of dual observation, which involves using a deep neural network with observation mechanisms to identify the main focus of a news article as well as the focus of the user on the article. This is achieved by taking into account the user's belief network, which reflects their personal interests and biases. By considering both the content of the news and the user's perspective, our approach is able to provide more personalised and accurate recommendations. We evaluate the performance of our model on real-world datasets and show that our proposed method outperforms several popular baselines.

cs.AI↗