arXiv subjects

Ruidong Wang

Publications and source records attributed to Ruidong Wang.

15 recordsLinked to original sources

arXiv2025

ARGUS: Defending Against Multimodal Indirect Prompt Injection via Steering Instruction-Following Behavior

Multimodal Large Language Models (MLLMs) are increasingly vulnerable to multimodal Indirect Prompt Injection (IPI) attacks, which embed malicious instructions in images, videos, or audio to hijack model behavior. Existing defenses, designed primarily for text-only LLMs, are unsuitable for countering these multimodal threats, as they are easily bypassed, modality-dependent, or generalize poorly. Inspired by activation steering researches, we hypothesize that a robust, general defense independent of modality can be achieved by steering the model's behavior in the representation space. Through extensive experiments, we discover that the instruction-following behavior of MLLMs is encoded in a subspace. Steering along directions within this subspace can enforce adherence to user instructions, forming the basis of a defense. However, we also found that a naive defense direction could be coupled with a utility-degrading direction, and excessive intervention strength harms model performance. To address this, we propose ARGUS, which searches for an optimal defense direction within the safety subspace that decouples from the utility degradation direction, further combining adaptive strength steering to achieve a better safety-utility trade-off. ARGUS also introduces lightweight injection detection stage to activate the defense on-demand, and a post-filtering stage to verify defense success. Experimental results show that ARGUS can achieve robust defense against multimodal IPI while maximally preserving the MLLM's utility.

cs.CR↗

arXiv2025

Efficient Reasoning via Chain of Unconscious Thought

Large Reasoning Models (LRMs) achieve promising performance but compromise token efficiency due to verbose reasoning processes. Unconscious Thought Theory (UTT) posits that complex problems can be solved more efficiently through internalized cognitive processes. Inspired by UTT, we propose a new reasoning paradigm, termed Chain of Unconscious Thought (CoUT), to improve the token efficiency of LRMs by guiding them to mimic human unconscious thought and internalize reasoning processes. Concretely, we first prompt the model to internalize the reasoning by thinking in the hidden layer. Then, we design a bag of token-efficient strategies to further help models reduce unnecessary tokens yet preserve the performance. Our work reveals that models may possess beneficial unconscious thought, enabling improved efficiency without sacrificing performance. Extensive experiments demonstrate the effectiveness of CoUT. Remarkably, it surpasses CoT by reducing token usage by 47.62% while maintaining comparable accuracy, as shown in Figure 1. The code of CoUT is available at this link: https://github.com/Rohan-GRH/CoUT

cs.CL↗

arXiv2024

Class-aware and Augmentation-free Contrastive Learning from Label Proportion

Learning from Label Proportion (LLP) is a weakly supervised learning scenario in which training data is organized into predefined bags of instances, disclosing only the class label proportions per bag. This paradigm is essential for user modeling and personalization, where user privacy is paramount, offering insights into user preferences without revealing individual data. LLP faces a unique difficulty: the misalignment between bag-level supervision and the objective of instance-level prediction, primarily due to the inherent ambiguity in label proportion matching. Previous studies have demonstrated deep representation learning can generate auxiliary signals to promote the supervision level in the image domain. However, applying these techniques to tabular data presents significant challenges: 1) they rely heavily on label-invariant augmentation to establish multi-view, which is not feasible with the heterogeneous nature of tabular datasets, and 2) tabular datasets often lack sufficient semantics for perfect class distinction, making them prone to suboptimality caused by the inherent ambiguity of label proportion matching. To address these challenges, we propose an augmentation-free contrastive framework TabLLP-BDC that introduces class-aware supervision (explicitly aware of class differences) at the instance level. Our solution features a two-stage Bag Difference Contrastive (BDC) learning mechanism that establishes robust class-aware instance-level supervision by disassembling the nuance between bag label proportions, without relying on augmentations. Concurrently, our model presents a pioneering multi-task pretraining pipeline tailored for tabular-based LLP, capturing intrinsic tabular feature correlations in alignment with label proportion distribution. Extensive experiments demonstrate that TabLLP-BDC achieves state-of-the-art performance for LLP in the tabular domain.

cs.LG↗

arXiv2023

HouYi: An open-source large language model specially designed for renewable energy and carbon neutrality field

Renewable energy is important for achieving carbon neutrality goal. With the great success of Large Language Models (LLMs) like ChatGPT in automatic content generation, LLMs are playing an increasingly important role. However, there has not been a specially designed LLM for renewable energy. Meanwhile, there has not been any dataset of renewable energy for training LLMs. Therefore, this paper published the first open-source Renewable Energy Academic Paper (REAP) dataset for non-commercial LLM research of renewable energy. REAP dataset is collected through searching the title and abstract of 1,168,970 academic literatures from Web of Science. Based on REAP dataset, HouYi model, the first LLM for renewable energy, is developed through finetuning general LLMs. HouYi demonstrated powerful academic paper paragraph generation ability in renewable energy field. Experiments show that its ability to generate academic papers on renewable energy is comparable to ChatGPT, slightly outperforms Claude, ERNIE Bot and SparkDesk, and significantly outperforms open-source LLaMA-13B model.

cs.CL↗

arXiv2021

Deep Dual Support Vector Data Description for Anomaly Detection on Attributed Networks

Networks are ubiquitous in the real world such as social networks and communication networks, and anomaly detection on networks aims at finding nodes whose structural or attributed patterns deviate significantly from the majority of reference nodes. However, most of the traditional anomaly detection methods neglect the relation structure information among data points and therefore cannot effectively generalize to the graph structure data. In this paper, we propose an end-to-end model of Deep Dual Support Vector Data description based Autoencoder (Dual-SVDAE) for anomaly detection on attributed networks, which considers both the structure and attribute for attributed networks. Specifically, Dual-SVDAE consists of a structure autoencoder and an attribute autoencoder to learn the latent representation of the node in the structure space and attribute space respectively. Then, a dual-hypersphere learning mechanism is imposed on them to learn two hyperspheres of normal nodes from the structure and attribute perspectives respectively. Moreover, to achieve joint learning between the structure and attribute of the network, we fuse the structure embedding and attribute embedding as the final input of the feature decoder to generate the node attribute. Finally, abnormal nodes can be detected by measuring the distance of nodes to the learned center of each hypersphere in the latent structure space and attribute space respectively. Extensive experiments on the real-world attributed networks show that Dual-SVDAE consistently outperforms the state-of-the-arts, which demonstrates the effectiveness of the proposed method.

cs.LG↗

arXiv2020

Correlation-aware Deep Generative Model for Unsupervised Anomaly Detection

Unsupervised anomaly detection aims to identify anomalous samples from highly complex and unstructured data, which is pervasive in both fundamental research and industrial applications. However, most existing methods neglect the complex correlation among data samples, which is important for capturing normal patterns from which the abnormal ones deviate. In this paper, we propose a method of Correlation aware unsupervised Anomaly detection via Deep Gaussian Mixture Model (CADGMM), which captures the complex correlation among data points for high-quality low-dimensional representation learning. Specifically, the relations among data samples are correlated firstly in forms of a graph structure, in which, the node denotes the sample and the edge denotes the correlation between two samples from the feature space. Then, a dual-encoder that consists of a graph encoder and a feature encoder, is employed to encode both the feature and correlation information of samples into the low-dimensional latent space jointly, followed by a decoder for data reconstruction. Finally, a separate estimation network as a Gaussian Mixture Model is utilized to estimate the density of the learned latent vector, and the anomalies can be detected by measuring the energy of the samples. Extensive experiments on real-world datasets demonstrate the effectiveness of the proposed method.

cs.LG↗

arXiv2020

On compactness and fixed point theorems in partial metric spaces

In this paper we examine two basic topological properties of partial metric spaces, namely compactness and completeness. Our main result claims that in these spaces compactness is equivalent to sequential compactness. We also show that Hausdorff compact partial metric spaces are metrizable. In the second part of this article we discuss the significance of bottom sets of partial metric spaces in fixed point theorems for mappings acting in these spaces.

math.GN↗

arXiv2020

Random bipartite posets and extremal problems

Previously, Erdős, Kierstead and Trotter investigated the dimension of random height~$2$ partially ordered sets. Their research was motivated primarily by two goals: (1)~analyzing the relative tightness of the Füredi-Kahn upper bounds on dimension in terms of maximum degree; and (2)~developing machinery for estimating the expected dimension of a random labeled poset on $n$ points. For these reasons, most of their effort was focused on the case $0<p\le 1/2$. While bounds were given for the range $1/2\le p <1$, the relative accuracy of the results in the original paper deteriorated as $p$ approaches~$1$. Motivated by two extremal problems involving conditions that force a poset to contain a large standard example, we were compelled to revisit this subject, but now with primary emphasis on the range $1/2\le p<1$. Our sharpened analysis shows that as $p$ approaches~$1$, the expected value of dimension increases and then decreases, answering in the negative a question posed in the original paper. Along the way, we apply inequalities of Talagrand and Janson, establish connections with latin rectangles and the Euler product function, and make progress on both extremal problems.

math.CO↗

arXiv2019

Planar Posets that are Accessible from Below Have Dimension at Most 6

Planar posets can have arbitrarily large dimension. However, a planar poset of height $h$ has dimension at most $192h+96$, while a planar poset with $t$ minimal elements has dimension at most $2t+1$. In particular, a planar poset with a unique minimal element has dimension at most $3$. In this paper, we extend this result by showing that a planar poset has dimension at most $6$ if it has a plane diagram in which every minimal element is accessible from below.

math.CO↗

arXiv2018

Dimension of posets with planar cover graphs excluding two long incomparable chains

It has been known for more than 40 years that there are posets with planar cover graphs and arbitrarily large dimension. Recently, Streib and Trotter proved that such posets must have large height. In fact, all known constructions of such posets have two large disjoint chains with all points in one chain incomparable with all points in the other. Gutowski and Krawczyk conjectured that this feature is necessary. More formally, they conjectured that for every $k\geq 1$, there is a constant $d$ such that if $P$ is a poset with a planar cover graph and $P$ excludes $\mathbf{k}+\mathbf{k}$, then $\dim(P)\leq d$. We settle their conjecture in the affirmative. We also discuss possibilities of generalizing the result by relaxing the condition that the cover graph is planar.

math.CO↗

arXiv2018

Dimension and cut vertices: an application of Ramsey theory

Motivated by quite recent research involving the relationship between the dimension of a poset and graph-theoretic properties of its cover graph, we show that for every $d\geq 1$, if $P$ is a poset and the dimension of a subposet $B$ of $P$ is at most $d$ whenever the cover graph of $B$ is a block of the cover graph of $P$, then the dimension of $P$ is at most $d+2$. We also construct examples which show that this inequality is best possible. We consider the proof of the upper bound to be fairly elegant and relatively compact. However, we know of no simple proof for the lower bound, and our argument requires a powerful tool known as the Product Ramsey Theorem. As a consequence, our constructions involve posets of enormous size.

math.CO↗

arXiv2016

On the dimension of posets with cover graphs of treewidth $2$

In 1977, Trotter and Moore proved that a poset has dimension at most $3$ whenever its cover graph is a forest, or equivalently, has treewidth at most $1$. On the other hand, a well-known construction of Kelly shows that there are posets of arbitrarily large dimension whose cover graphs have treewidth $3$. In this paper we focus on the boundary case of treewidth $2$. It was recently shown that the dimension is bounded if the cover graph is outerplanar (Felsner, Trotter, and Wiechert) or if it has pathwidth $2$ (Biró, Keller, and Young). This can be interpreted as evidence that the dimension should be bounded more generally when the cover graph has treewidth $2$. We show that it is indeed the case: Every such poset has dimension at most $1276$.

math.CO↗

arXiv2015

Tree-width and dimension

Over the last 30 years, researchers have investigated connections between dimension for posets and planarity for graphs. Here we extend this line of research to the structural graph theory parameter tree-width by proving that the dimension of a finite poset is bounded in terms of its height and the tree-width of its cover graph.

math.CO↗

arXiv2013

Incidence Posets and Cover Graphs

We prove two theorems concerning incidence posets of graphs, cover graphs of posets and a related graph parameter. First, answering a question of Haxell, we show that the chromatic number of a graph is not bounded in terms of the dimension of its incidence poset, provided the dimension is at least four. Second, answering a question of Kříž and Nešetřil, we show that there are graphs with large girth and large chromatic number among the class of graphs having eye parameter at most two.

math.CO↗

arXiv2012

On the numerical radius of Lipschitz operators in Banach spaces

We study the numerical radius of Lipschitz operators on Banach spaces via the Lipschitz numerical index, which is an analogue of the numerical index in Banach space theory. We give a characterization of the numerical radius and obtain a necessary and sufficient condition for Banach spaces to have Lipschitz numerical index 1. As an application, we show that real lush spaces and $C$-rich subspaces have Lipschitz numerical index 1. Moreover, using the G$\hat{a}$teaux differentiability of Lipschitz operators, we characterize the Lipschitz numerical index of separable Banach spaces with the RNP. Finally, we prove that the Lipschitz numerical index has the stability properties for the $c_0$-, $l_1$-, and $l_\infty$-sums of spaces and vector-valued function spaces. From this, we show that the $C(K)$ spaces, $L_1(μ)$-spaces and $L_\infty(ν)$ spaces have Lipschitz numerical index 1.

math.FA↗