Author name cluster

Kun Zhu

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

10 papers

1 author row

EAAI Journal 2026 Journal Article

Diversity-driven reasoning: Mitigating logical errors in LLMs through social-attribute guided multi-agent collaboration

Liangji Zhang
Jianbo Yuan
Yougming He
Miao Yu
Kun Zhu
Zhenni Yu

This paper proposes a Multi-Attribute Hybrid Expert (MAME) framework to address the challenges of error solidification in large language models (LLMs) arising from logical error accumulation and answer convergence during multi-round complex reasoning tasks. First, we introduce a Group Feedback Framework (GFF) that integrates group decision-making theory into multi-agent collaboration. By simulating human consensus formation and error correction mechanisms, we define a dynamic confidence matrix and an answer credibility score to quantify inter-agent influence. These metrics dynamically guide agents in adjusting their reasoning paths based on collective feedback. Second, we propose a Multi-Attribute Prompting (MAP) method that assigns distinct social attributes to each agent, fostering perspective diversity through weight vector modulation and generation strategies. This approach effectively mitigates logical error accumulation, significantly suppresses error propagation across rounds, and prevents answer convergence and solidification. We evaluated MAME using Qwen2. 5 models at scales of 3B, 7B, 14B, 32B, and 72B parameters. Experimental results demonstrate significant performance improvements across reasoning tasks: MAME achieves enhancements of up to 24. 05 % in mathematical reasoning, 39. 1 % in logical reasoning, and 4. 94 % in commonsense reasoning compared to baseline models. Notably, MAME represents the first framework to embed social personality theory within multi-agent cooperation, providing an interpretable error correction mechanism for LLM-based complex reasoning. Our findings highlight the critical role of personality diversity in mitigating logical stagnation, laying a foundation for the deeper integration of LLMs with behavioral science and decision theory. The code is open source: https: //github. com/ioio0614/MAME.

Details DOI

AAAI Conference 2026 Conference Paper

Targeting Borderline Fraudsters: Multi-View Hypergraph Fraud Detection with LLM-Guided Contrastive Learning

Rui Ou
Kun Zhu
Nana Zhang
Jiangtong Li
Chaochao Chen
Yuhua Xu
Changjun Jiang

Graph fraud detection (GFD) on transaction networks is crucial for safeguarding financial systems. However, due to the limited perspective of existing graph neural networks (GNNs) in the single transaction view, sophisticated fraudsters can disguise themselves to exhibit weak fraud signals, appearing as borderline fraudsters. To address this challenge, we propose MH-LGC, a multi-view hypergraph fraud detection model with large language model (LLM) guided contrastive learning. MH-LGC tackles two key limitations of existing GNN-based GFD methods: (1) Due to the local aggregation mechanism, existing methods struggle to capture high-order trading patterns among distant fraudsters. MH-LGC introduces two temporal hyper-views as complements to the transaction view and employs a Temporal Hypergraph Attention Network (THAN) to integrate the three views. (2) Most GFD methods overlook the rich semantic cues embedded in transaction data. Although some general graph learning studies have explored LLM integration, the high computational overhead and task-specific fine-tuning make them impractical for GFD tasks. MH-LGC introduces a semantic view through a fine-tuning-free LLM-Guided Contrastive learning (LGC), adopting a novel paradigm for integrating GNN and LLM to reduce the computational overhead of LLM. Extensive experiments on three real-world datasets demonstrate that MH-LGC outperforms twelve state-of-the-art baselines, with AUC improvements ranging from 1.10% to 5.70%.

PDF Details DOI

NeurIPS Conference 2025 Conference Paper

Attack by Yourself: Effective and Unnoticeable Multi-Category Graph Backdoor Attacks with Subgraph Triggers Pool

Jiangtong Li
Dongyi Liu
Kun Zhu
Dawei Cheng
Changjun Jiang

Graph Neural Networks (GNNs) have achieved significant success in various real-world applications, including social networks, finance systems, and traffic management. Recent researches highlight their vulnerability to backdoor attacks in node classification, where GNNs trained on a poisoned graph misclassify a test node only when specific triggers are attached. These studies typically focus on single attack categories and use adaptive trigger generators to create node-specific triggers. However, adaptive trigger generators typically have a simple structure, limited parameters, and lack category-aware graph knowledge, which makes them struggle to handle backdoor attacks across multiple categories as the number of target categories increases. We address this gap by proposing a novel approach for Effective and Unnoticeable Multi-Category (EUMC) graph backdoor attacks, leveraging subgraph from the attacked graph as category-aware triggers to precisely control the target category. To ensure the effectiveness of our method, we construct a Multi-Category Subgraph Triggers Pool (MC-STP) using the subgraphs of the attacked graph as triggers. We then exploit the attachment probability shifts of each subgraph trigger as category-aware priors for target category determination. Moreover, we develop a ``select then attach'' strategy that connects suitable category-aware trigger to attacked nodes for unnoticeability. Extensive experiments across different real-world datasets confirm the efficacy of our method in conducting multi-category graph backdoor attacks on various GNN models and defense strategies.

PDF Details

NeurIPS Conference 2025 Conference Paper

Encouraging metric-aware diversity in contrastive representation space

Tianxu Li
Kun Zhu

In cooperative Multi-Agent Reinforcement Learning (MARL), agents that share policy network parameters often learn similar behaviors, which hinders effective exploration and can lead to suboptimal cooperative policies. Recent advances have attempted to promote multi-agent diversity by leveraging the Wasserstein distance to increase policy differences. However, these methods cannot effectively encourage diverse policies due to ineffective Wasserstein distance caused by the policy similarity. To address this limitation, we propose Wasserstein Contrastive Diversity (WCD) exploration, a novel approach that promotes multi-agent diversity by maximizing the Wasserstein distance between the trajectory distributions of different agents in a latent representation space. To make the Wasserstein distance meaningful, we propose a novel next-step prediction method based on Contrastive Predictive Coding (CPC) to learn distinguishable trajectory representations. Additionally, we introduce an optimized kernel-based method to compute the Wasserstein distance more efficiently. Since the Wasserstein distance is inherently defined for two distributions, we extend it to support multiple agents, enabling diverse policy learning. Empirical evaluations across a variety of challenging multi-agent tasks demonstrate that WCD outperforms existing state-of-the-art methods, delivering superior performance and enhanced exploration.

PDF Details

AAAI Conference 2025 Conference Paper

Learning Joint Behaviors with Large Variations

Tianxu Li
Kun Zhu

Cooperative Multi-Agent Reinforcement Learning (MARL) has drawn increasing interest in recent works due to its significant achievements. However, there are still some challenges impeding the learning of optimal cooperative policies, such as insufficient exploration. Prior works typically adopt mutual information-based methods to encourage exploration. However, this category of methods does not necessarily encourage agents to fully explore the joint behavior space. To address this limitation, we propose a novel objective based on learning a representation function with a Lipschitz constraint to maximize the traveled distances in the joint behavior space, encouraging agents to learn joint behaviors with large variations and leading to sufficient exploration. We further implement our method on top of QMIX. We demonstrate the effectiveness of our method by conducting experiments on the LBF, SMAC, and SMACv2 benchmarks. Our method outperforms previous methods in terms of final performance and state-action space exploration.

PDF Details DOI

AAMAS Conference 2025 Conference Paper

Self-Supervised Multi-Agent Diversity with Nonparametric Entropy Maximization

Tianxu Li
Kun Zhu

Learning decentralized policies for agents has drawn increasing interest in recent works to solve the scalability issue that arises in Multi-Agent Reinforcement Learning (MARL), where all agents may share the parameters of a policy network to make action decisions. However, such parameter sharing can hinder efficient exploration, as some agents may learn similar behaviors. Unlike previous fullysupervised mutual information-based methods that encourages multi-agent diversity, in this paper, we propose a novel multi-agent exploration method called Contrastive Trajectory Entropy Maximization (CTEM). Our method adopts a non-parametric entropy estimator to maximize the entropy of trajectories of different agents in a self-supervised contrastive representation space, leading to diverse policies and sufficient exploration. Such an entropy estimator avoids complex density modeling and scales well in highdimensional multi-agent environments. We deploy our method in MARL by introducing an intrinsic reward for agents to achieve entropy maximization. To demonstrate the effectiveness of our method, we conduct experiments on multiple challenging MARL benchmark tasks. Our method yields superior performance than existing state-of-the-art methods.

PDF

NeurIPS Conference 2024 Conference Paper

Discrete Modeling via Boundary Conditional Diffusion Processes

Yuxuan Gu
Xiaocheng Feng
Lei Huang
Yingsheng Wu
Zekun Zhou
Weihong Zhong
Kun Zhu
Bing Qin

We present an novel framework for efficiently and effectively extending the powerful continuous diffusion processes to discrete modeling. Previous approaches have suffered from the discrepancy between discrete data and continuous modeling. Our study reveals that the absence of guidance from discrete boundaries in learning probability contours is one of the main reasons. To address this issue, we propose a two-step forward process that first estimates the boundary as a prior distribution and then rescales the forward trajectory to construct a boundary conditional diffusion model. The reverse process is proportionally adjusted to guarantee that the learned contours yield more precise discrete data. Experimental results indicate that our approach achieves strong performance in both language modeling and discrete image generation tasks. In language modeling, our approach surpasses previous state-of-the-art continuous diffusion language models in three translation tasks and a summarization task, while also demonstrating competitive performance compared to auto-regressive transformers. Moreover, our method achieves comparable results to continuous diffusion models when using discrete ordinal pixels and establishes a new state-of-the-art for categorical image generation on the Cifar-10 dataset.

PDF Details DOI

NeurIPS Conference 2024 Conference Paper

Learning Distinguishable Trajectory Representation with Contrastive Loss

Tianxu Li
Kun Zhu
Juan Li
Yang Zhang

Policy network parameter sharing is a commonly used technique in advanced deep multi-agent reinforcement learning (MARL) algorithms to improve learning efficiency by reducing the number of policy parameters and sharing experiences among agents. Nevertheless, agents that share the policy parameters tend to learn similar behaviors. To encourage multi-agent diversity, prior works typically maximize the mutual information between trajectories and agent identities using variational inference. However, this category of methods easily leads to inefficient exploration due to limited trajectory visitations. To resolve this limitation, inspired by the learning of pre-trained models, in this paper, we propose a novel Contrastive Trajectory Representation (CTR) method based on learning distinguishable trajectory representations to encourage multi-agent diversity. Specifically, CTR maps the trajectory of an agent into a latent trajectory representation space by an encoder and an autoregressive model. To achieve the distinguishability among trajectory representations of different agents, we introduce contrastive learning to maximize the mutual information between the trajectory representations and learnable identity representations of different agents. We implement CTR on top of QMIX and evaluate its performance in various cooperative multi-agent tasks. The empirical results demonstrate that our proposed CTR yields significant performance improvement over the state-of-the-art methods.

PDF Details DOI

YNIMG Journal 2024 Journal Article

The enhanced connectivity between the frontoparietal, somatomotor network and thalamus as the most significant network changes of chronic low back pain

Kun Zhu
Jianchao Chang
Siya Zhang
Yan Li
Junxun Zuo
Haoyu Ni
Bingyong Xie
Jiyuan Yao

The prolonged duration of chronic low back pain (cLBP) inevitably leads to changes in the cognitive, attentional, sensory and emotional processing brain regions. Currently, it remains unclear how these alterations are manifested in the interplay between brain functional and structural networks. This study aimed to predict the Oswestry Disability Index (ODI) in cLBP patients using multimodal brain magnetic resonance imaging (MRI) data and identified the most significant features within the multimodal networks to aid in distinguishing patients from healthy controls (HCs). We constructed dynamic functional connectivity (dFC) and structural connectivity (SC) networks for all participants (n = 112) and employed the Connectome-based Predictive Modeling (CPM) approach to predict ODI scores, utilizing various feature selection thresholds to identify the most significant network change features in dFC and SC outcomes. Subsequently, we utilized these significant features for optimal classifier selection and the integration of multimodal features. The results revealed enhanced connectivity among the frontoparietal network (FPN), somatomotor network (SMN) and thalamus in cLBP patients compared to HCs. The thalamus transmits pain-related sensations and emotions to the cortical areas through the dorsolateral prefrontal cortex (dlPFC) and primary somatosensory cortex (SI), leading to alterations in whole-brain network functionality and structure. Regarding the model selection for the classifier, we found that Support Vector Machine (SVM) best fit these significant network features. The combined model based on dFC and SC features significantly improved classification performance between cLBP patients and HCs (AUC=0.9772). Finally, the results from an external validation set support our hypotheses and provide insights into the potential applicability of the model in real-world scenarios. Our discovery of enhanced connectivity between the thalamus and both the dlPFC (FPN) and SI (SMN) provides a valuable supplement to prior research on cLBP.

Details DOI

AAAI Conference 2024 Conference Paper

Towards the Disappearing Truth: Fine-Grained Joint Causal Influences Learning with Hidden Variable-Driven Causal Hypergraphs in Time Series

Kun Zhu
Chunhui Zhao

Causal discovery under Granger causality framework has yielded widespread concerns in time series analysis task. Nevertheless, most previous methods are unaware of the underlying causality disappearing problem, that is, certain weak causalities are less focusable and may be lost during the modeling process, thus leading to biased causal conclusions. Therefore, we propose to introduce joint causal influences (i.e., causal influences from the union of multiple variables) as additional causal indication information to help identify weak causalities. Further, to break the limitation of existing methods that implicitly and coarsely model joint causal influences, we propose a novel hidden variable-driven causal hypergraph neural network to meticulously explore the locality and diversity of joint causal influences, and realize its explicit and fine-grained modeling. Specifically, we introduce hidden variables to construct a causal hypergraph for explicitly characterizing various fine-grained joint causal influences. Then, we customize a dual causal information transfer mechanism (encompassing a multi-level causal path and an information aggregation path) to realize the free diffusion and meticulous aggregation of joint causal influences and facilitate its adaptive learning. Finally, we design a multi-view collaborative optimization constraint to guarantee the characterization diversity of causal hypergraph and capture remarkable forecasting relationships (i.e., causalities). Experiments are conducted to demonstrate the superiority of the proposed model.

PDF Details DOI