arXiv subjects

Guangyu He

Publications and source records attributed to Guangyu He.

6 recordsLinked to original sources

arXiv2026

Null Space Constrained Contrastive Visual Forgetting for MLLM Unlearning

The core challenge of machine unlearning is to strike a balance between target knowledge removal and non-target knowledge retention. In the context of Multimodal Large Language Models (MLLMs), this challenge becomes even more pronounced, as knowledge is further divided into visual and textual modalities that are tightly intertwined. In this paper, we introduce an MLLM unlearning approach that aims to forget target visual knowledge while preserving non-target visual knowledge and all textual knowledge. Specifically, we freeze the LLM backbone and achieve unlearning by fine-tuning the visual module. First, we propose a Contrastive Visual Forgetting (CVF) mechanism to separate target visual knowledge from retained visual knowledge, guiding the representations of target visual concepts toward appropriate regions in the feature space. Second, we identify the null space associated with retained knowledge and constrain the unlearning process within this space, thereby significantly mitigating degradation in knowledge retention. Third, beyond static unlearning scenarios, we extend our approach to continual unlearning, where forgetting requests arrive sequentially. Extensive experiments across diverse benchmarks demonstrate that our approach achieves a strong balance between effective forgetting and robust knowledge retention.

cs.AI↗

arXiv2026

ICU-Bench:Benchmarking Continual Unlearning in Multimodal Large Language Models

Privacy deletion requests often arrive sequentially, creating a continual unlearning challenge for deployed multimodal large language models (MLLMs). However, existing benchmarks mainly focus on static or short-sequence settings, offering limited support for evaluating continual privacy deletion on privacy-critical documents. To bridge this gap, we introduce ICU-Bench, an Identity-centric Continual Unlearning benchmark (pronounced ``I see you'') for privacy-critical multimodal documents. ICU-Bench contains 1,000 synthetic privacy-sensitive profiles from medical reports and labor contracts, comprising 9,500 document images, 16,000 question-answer pairs, and 100 sequential forget tasks. We further introduce history-aware evaluation protocols and sequence-aware metrics to assess current forgetting, historical forgetting preservation, retained utility, and stability throughout the unlearning sequence. Experiments with representative unlearning methods on two MLLMs reveal that methods effective on current targets often fail to preserve earlier forgetting or retained capabilities over long sequences. Some methods further obtain low forget accuracy through severe model degradation. These results expose long-horizon failure modes overlooked by conventional evaluations and highlight the need for multimodal unlearning methods explicitly designed for continual privacy deletion. Our code and benchmark are publicly available at: https://github.com/AstorYH/ICU-Bench.

cs.AI↗

arXiv2026

From Assistant to Double Agent: Formalizing and Benchmarking Attacks on OpenClaw for Personalized Local AI Agent

Although large language model (LLM)-based agents, exemplified by OpenClaw, are increasingly evolving from task-oriented systems into personalized AI assistants for solving complex real-world tasks, their practical deployment also introduces severe security risks. However, existing agent security research and evaluation frameworks primarily focus on synthetic or task-centric settings, and thus fail to accurately capture the attack surface and risk propagation mechanisms of personalized agents in real-world deployments. To address this gap, we propose Personalized Agent Security Bench (PASB), an end-to-end security evaluation framework tailored for real-world personalized agents. Building upon existing agent attack paradigms, PASB incorporates personalized usage scenarios, realistic toolchains, and long-horizon interactions, enabling black-box, end-to-end security evaluation on real systems. Using OpenClaw as a representative case study, we systematically evaluate its security across multiple personalized scenarios, tool capabilities, and attack types. Our results indicate that OpenClaw exhibits critical vulnerabilities at different execution stages, including user prompt processing, tool usage, and memory retrieval, highlighting substantial security risks in personalized agent deployments. The code for the proposed PASB framework is available at https://github.com/AstorYH/PASB.

cs.AI↗

arXiv2025

Robust MLLM Unlearning via Visual Knowledge Distillation

Recently, machine unlearning approaches have been proposed to remove sensitive information from well-trained large models. However, most existing methods are tailored for LLMs, while MLLM-oriented unlearning remains at its early stage. Inspired by recent studies exploring the internal mechanisms of MLLMs, we propose to disentangle the visual and textual knowledge embedded within MLLMs and introduce a dedicated approach to selectively erase target visual knowledge while preserving textual knowledge. Unlike previous unlearning methods that rely on output-level supervision, our approach introduces a Visual Knowledge Distillation (VKD) scheme, which leverages intermediate visual representations within the MLLM as supervision signals. This design substantially enhances both unlearning effectiveness and model utility. Moreover, since our method only fine-tunes the visual components of the MLLM, it offers significant efficiency advantages. Extensive experiments demonstrate that our approach outperforms state-of-the-art unlearning methods in terms of both effectiveness and efficiency. Moreover, we are the first to evaluate the robustness of MLLM unlearning against relearning attacks.

cs.CV↗

arXiv2022

Machine Learning Enabled Force Sensing of a Smart Skin for Robotics

Artificial skin with the sense of touch can support robots to interact with the surrounding environment efficiently and accomplish complex tasks. Traditional multi-layered artificial skins require complex manufacturing processes which can result in high cost as well as limitations on the material and size of the skin. In this paper, we demonstrate a machine learning based approach to predict positions of point loads using the most direct response as input signal: strain distribution. Starting with the simplest problem, predicting the position of a single point load acting on a flat surface, an ML model is developed, trained, and tested. Accurate predictions are obtained from the ML model, parameters that affect the accuracy are discussed, and validation tests are performed. After that, the ML model is modified to solve multi-objective prediction problems: predicting positions and magnitudes of multiple point loads. In the end, the ML model is upgraded to a 2-step model to predict the position of a point load acting on a deformable surface. The demonstrated approach enables a normal untreated surface to feel a touch no matter what the surface is made of or how large or small the size of the surface is. Therefore, we believe this ML-based load position prediction approach could be a promising tool for applications such as flexible touch screens, smart skin for robots, and micro touch sensors.

physics.app-ph↗

arXiv2017

Impacts and Benefits of UPFC to Wind Power Integration in Unit Commitment

Unified Power Flow Controller (UPFC) is recognized as the most powerful flexible AC transmission systems (FACTS) device for power system operation. This paper addresses how UPFC explores the transmission flexibility and facilitates the integration of uncertain and volatile wind power generation. To this end, a comprehensive unit commitment (UC) model with UPFC and uncertain wind power generation is proposed. Then, some metrics are introduced to evaluate the impacts of UPFC on the reliability, security and economy of power system operation. Further, different dispatch strategies of UPFC are compared to provide helpful guidances on making full use of UPFC to hedge against uncertainties. In addition, facing the challenging mixed-integer non-linear non-convex problems, approximate models are proposed to provide a starting point to solve the problems efficiently. All these models are easy to adapt to other types of FACTS devices. Illustrative numerical results are provided.

math.OC↗