Zhixue Song

Zhixue Song, Boyan Han, Yiwei Wang et al.

Recent advancements in visual context compression enable MLLMs to process ultra-long contexts efficiently by rendering text into images. However, we identify a critical vulnerability inherent to this paradigm: lowering image resolution inadvertently catalyzes jailbreaking. Our experiments reveal that the safety defenses of SOTA models deteriorate sharply as resolution degrades, surprisingly persisting even when text remains legible. We attribute this to ``Cognitive Overload'', hypothesizing that the effort required to decipher degraded inputs diverts attentional resources from safety auditing. This phenomenon is consistent across various visual perturbations, including noise and geometric distortion. To address this, we propose a simple ``Structured Cognitive Offloading'' strategy that mitigates these risks by enforcing a serialized pipeline to decouple visual transcription from safety assessment. Our work exposes a significant risk in vision-based compression and provides critical insights for the secure design of future MLLMs.

Zhuoheng Li, Yaochen Wang, Zhixue Song et al.

This study explores the capabilities of large language models (LLMs) in providing knowledge about cities and regions on a global scale. We employ two methods: directly querying the LLM for target variable values and extracting explicit and implicit features from the LLM correlated with the target variable. Our experiments reveal that LLMs embed a broad but varying degree of knowledge across global cities, with ML models trained on LLM-derived features consistently leading to improved predictive accuracy. Additionally, we observe that LLMs demonstrate a certain level of knowledge across global cities on all continents, but it is evident when they lack knowledge, as they tend to generate generic or random outputs for unfamiliar tasks. These findings suggest that LLMs can offer new opportunities for data-driven decision-making in the study of cities.

Zhixue Song, Zhiheng Zhang, Yi Song et al.

The OpenClaw platform provides a practical foundation for automation through its skill-oriented architecture, organizing external capabilities into lightweight, reusable components that can be invoked efficiently through a command-line interface (CLI). However, a significant bottleneck remains: many real-world tasks are confined to graphical user interfaces (GUIs) with no stable API available. While LLM-based GUI agents offer generality, their reliance on repeated live model inference makes them too slow, costly, and inconsistent to serve as efficient OpenClaw skills. In this paper, we present AppAgent-Claw, a demonstration-driven system that converts GUI workflows into reliable, reusable skills without runtime inference. By following a ``record-once, replay-many'' paradigm, the system captures rich contextual metadata to facilitate robust execution. It employs a layered localization strategy to handle visual shifts and a validation-coupled execution model to ensure intended on-screen effects. AppAgent-Claw provides a practical, efficient, and diagnosable solution for integrating GUI-bound tasks into the OpenClaw ecosystem.