arXiv subjects

Litian Zhang

Publications and source records attributed to Litian Zhang.

At least 37 records · Page 2Linked to original sources

arXiv2026

From static to adaptive: immune memory-based jailbreak detection for large language models

Large Language Models (LLMs) serve as the backbone of modern AI systems, yet they remain susceptible to adversarial jailbreak attacks. Consequently, robust detection of such malicious inputs is paramount for ensuring model safety. Traditional detection methods typically rely on external models trained on fixed, large-scale datasets, which often incur significant computational overhead. While recent methods shift toward leveraging internal safety signals of models to enable more lightweight and efficient detection. However, these methods remain inherently static and struggle to adapt to the evolving nature of jailbreak attacks. Drawing inspiration from the biological immune mechanism, we introduce the Immune Memory Adaptive Guard (IMAG) framework. By distilling and encoding safety patterns into a persistent, evolvable memory bank, IMAG enables adaptive generalization to emerging threats. Specifically, the framework orchestrates three synergistic components: Immune Detection, which employs retrieval for high-efficiency interception of known jailbreak attacks; Active Immunity, which performs proactive behavioral simulation to resolve ambiguous unknown queries; Memory Updating, which integrates validated attack patterns back into the memory bank. This closed-loop architecture transitions LLM defense from rigid filtering to autonomous adaptive mitigation. Extensive evaluations across five representative open-source LLMs demonstrate that our method surpasses state-of-the-art (SOTA) baselines, achieving a superior average detection accuracy of 94\% across diverse and complex attack types.

cs.CR↗

arXiv2026

Jailbreaking LLMs & VLMs: Mechanisms, Evaluation, and Unified Defense

This paper provides a systematic survey of jailbreak attacks and defenses on Large Language Models (LLMs) and Vision-Language Models (VLMs), emphasizing that jailbreak vulnerabilities stem from structural factors such as incomplete training data, linguistic ambiguity, and generative uncertainty. It further differentiates between hallucinations and jailbreaks in terms of intent and triggering mechanisms. We propose a three-dimensional survey framework: (1) Attack dimension-including template/encoding-based, in-context learning manipulation, reinforcement/adversarial learning, LLM-assisted and fine-tuned attacks, as well as prompt- and image-level perturbations and agent-based transfer in VLMs; (2) Defense dimension-encompassing prompt-level obfuscation, output evaluation, and model-level alignment or fine-tuning; and (3) Evaluation dimension-covering metrics such as Attack Success Rate (ASR), toxicity score, query/time cost, and multimodal Clean Accuracy and Attribute Success Rate. Compared with prior works, this survey spans the full spectrum from text-only to multimodal settings, consolidating shared mechanisms and proposing unified defense principles: variant-consistency and gradient-sensitivity detection at the perception layer, safety-aware decoding and output review at the generation layer, and adversarially augmented preference alignment at the parameter layer. Additionally, we summarize existing multimodal safety benchmarks and discuss future directions, including automated red teaming, cross-modal collaborative defense, and standardized evaluation.

cs.CR↗

arXiv2026

LANCET: Neural Intervention via Structural Entropy for Mitigating Faithfulness Hallucinations in LLMs

Large Language Models have revolutionized information processing, yet their reliability is severely compromised by faithfulness hallucinations. While current approaches attempt to mitigate this issue through node-level adjustments or coarse suppression, they often overlook the distributed nature of neural information, leading to imprecise interventions. Recognizing that hallucinations propagate through specific forward transmission pathways like an infection, we aim to surgically block this flow using precise structural analysis. To leverage this, we propose Lancet, a novel framework that achieves precise neural intervention by leveraging structural entropy and hallucination difference ratios. Lancet first locates hallucination-prone neurons via gradient-driven contrastive analysis, then maps their propagation pathways by minimizing structural entropy, and finally implements a hierarchical intervention strategy that preserves general model capabilities. Comprehensive evaluations across hallucination benchmark datasets demonstrate that Lancet significantly outperforms state-of-the-art methods, validating the effectiveness of our surgical approach to neural intervention.

cs.CL↗

arXiv2026

How Real is Your Jailbreak? Fine-grained Jailbreak Evaluation with Anchored Reference

Jailbreak attacks present a significant challenge to the safety of Large Language Models (LLMs), yet current automated evaluation methods largely rely on coarse classifications that focus mainly on harmfulness, leading to substantial overestimation of attack success. To address this problem, we propose FJAR, a fine-grained jailbreak evaluation framework with anchored references. We first categorized jailbreak responses into five fine-grained categories: Rejective, Irrelevant, Unhelpful, Incorrect, and Successful, based on the degree to which the response addresses the malicious intent of the query. This categorization serves as the basis for FJAR. Then, we introduce a novel harmless tree decomposition approach to construct high-quality anchored references by breaking down the original queries. These references guide the evaluator in determining whether the response genuinely fulfills the original query. Extensive experiments demonstrate that FJAR achieves the highest alignment with human judgment and effectively identifies the root causes of jailbreak failures, providing actionable guidance for improving attack strategies.

cs.CR↗

arXiv2025

One SPACE to Rule Them All: Jointly Mitigating Factuality and Faithfulness Hallucinations in LLMs

LLMs have demonstrated unprecedented capabilities in natural language processing, yet their practical deployment remains hindered by persistent factuality and faithfulness hallucinations. While existing methods address these hallucination types independently, they inadvertently induce performance trade-offs, as interventions targeting one type often exacerbate the other. Through empirical and theoretical analysis of activation space dynamics in LLMs, we reveal that these hallucination categories share overlapping subspaces within neural representations, presenting an opportunity for concurrent mitigation. To harness this insight, we propose SPACE, a unified framework that jointly enhances factuality and faithfulness by editing shared activation subspaces. SPACE establishes a geometric foundation for shared subspace existence through dual-task feature modeling, then identifies and edits these subspaces via a hybrid probe strategy combining spectral clustering and attention head saliency scoring. Experimental results across multiple benchmark datasets demonstrate the superiority of our approach.

cs.CL↗

arXiv2025

The Scales of Justitia: A Comprehensive Survey on Safety Evaluation of LLMs

With the rapid advancement of artificial intelligence, Large Language Models (LLMs) have shown remarkable capabilities in Natural Language Processing (NLP), including content generation, human-computer interaction, machine translation, and code generation. However, their widespread deployment has also raised significant safety concerns. In particular, LLM-generated content can exhibit unsafe behaviors such as toxicity, bias, or misinformation, especially in adversarial contexts, which has attracted increasing attention from both academia and industry. Although numerous studies have attempted to evaluate these risks, a comprehensive and systematic survey on safety evaluation of LLMs is still lacking. This work aims to fill this gap by presenting a structured overview of recent advances in safety evaluation of LLMs. Specifically, we propose a four-dimensional taxonomy: (i) Why to evaluate, which explores the background of safety evaluation of LLMs, how they differ from general LLMs evaluation, and the significance of such evaluation; (ii) What to evaluate, which examines and categorizes existing safety evaluation tasks based on key capabilities, including dimensions such as toxicity, robustness, ethics, bias and fairness, truthfulness, and related aspects; (iii) Where to evaluate, which summarizes the evaluation metrics, datasets and benchmarks currently used in safety evaluations; (iv) How to evaluate, which reviews existing mainstream evaluation methods based on the roles of the evaluators and some evaluation frameworks that integrate the entire evaluation pipeline. Finally, we identify the challenges in safety evaluation of LLMs and propose promising research directions to promote further advancement in this field. We emphasize the necessity of prioritizing safety evaluation to ensure the reliable and responsible deployment of LLMs in real-world applications.

cs.CL↗

arXiv2025

Beyond Surface-Level Detection: Towards Cognitive-Driven Defense Against Jailbreak Attacks via Meta-Operations Reasoning

Defending large language models (LLMs) against jailbreak attacks is essential for their safe and reliable deployment. Existing defenses often rely on shallow pattern matching, which struggles to generalize to novel and unseen attack strategies. To address this challenge, we propose the Cognitive-Driven Defense (CDD) framework, which targets the underlying structure of jailbreak prompts by applying meta-operations, defined as basic manipulations that conceal harmful intent.CDD emulates human cognitive reasoning through a structured reasoning chain. It begins with a global perception of the prompt and follows with a localized analysis to uncover hidden manipulations. By applying supervised fine-tuning on this structured chain, the model learns to identify and reason about known manipulation patterns. To enhance generalization to unseen threats, an entropy-guided reinforcement learning algorithm (EG-GRPO) is introduced to encourage exploration of new types and variants of meta-operations. Experiments demonstrate that CDD can achieve state-of-the-art defense performance and exhibit strong generalization to unseen jailbreak attacks.

cs.AI↗

arXiv2025

Attack the Messages, Not the Agents: A Multi-round Adaptive Stealthy Tampering Framework for LLM-MAS

Large language model-based multi-agent systems (LLM-MAS) effectively accomplish complex and dynamic tasks through inter-agent communication, but this reliance introduces substantial safety vulnerabilities. Existing attack methods targeting LLM-MAS either compromise agent internals or rely on direct and overt persuasion, which limit their effectiveness, adaptability, and stealthiness. In this paper, we propose MAST, a Multi-round Adaptive Stealthy Tampering framework designed to exploit communication vulnerabilities within the system. MAST integrates Monte Carlo Tree Search with Direct Preference Optimization to train an attack policy model that adaptively generates effective multi-round tampering strategies. Furthermore, to preserve stealthiness, we impose dual semantic and embedding similarity constraints during the tampering process. Comprehensive experiments across diverse tasks, communication architectures, and LLMs demonstrate that MAST consistently achieves high attack success rates while significantly enhancing stealthiness compared to baselines. These findings highlight the effectiveness, stealthiness, and adaptability of MAST, underscoring the need for robust communication safeguards in LLM-MAS.

cs.CR↗

arXiv2025

Feint and Attack: Attention-Based Strategies for Jailbreaking and Protecting LLMs

Jailbreak attack can be used to access the vulnerabilities of Large Language Models (LLMs) by inducing LLMs to generate the harmful content. And the most common method of the attack is to construct semantically ambiguous prompts to confuse and mislead the LLMs. To access the security and reveal the intrinsic relation between the input prompt and the output for LLMs, the distribution of attention weight is introduced to analyze the underlying reasons. By using statistical analysis methods, some novel metrics are defined to better describe the distribution of attention weight, such as the Attention Intensity on Sensitive Words (Attn_SensWords), the Attention-based Contextual Dependency Score (Attn_DepScore) and Attention Dispersion Entropy (Attn_Entropy). By leveraging the distinct characteristics of these metrics, the beam search algorithm and inspired by the military strategy "Feint and Attack", an effective jailbreak attack strategy named as Attention-Based Attack (ABA) is proposed. In the ABA, nested attack prompts are employed to divert the attention distribution of the LLMs. In this manner, more harmless parts of the input can be used to attract the attention of the LLMs. In addition, motivated by ABA, an effective defense strategy called as Attention-Based Defense (ABD) is also put forward. Compared with ABA, the ABD can be used to enhance the robustness of LLMs by calibrating the attention distribution of the input prompt. Some comparative experiments have been given to demonstrate the effectiveness of ABA and ABD. Therefore, both ABA and ABD can be used to access the security of the LLMs. The comparative experiment results also give a logical explanation that the distribution of attention weight can bring great influence on the output for LLMs.

cs.CR↗

arXiv2025

LLMs are Introvert

The exponential growth of social media and generative AI has transformed information dissemination, fostering connectivity but also accelerating the spread of misinformation. Understanding information propagation dynamics and developing effective control strategies is essential to mitigate harmful content. Traditional models, such as SIR, provide basic insights but inadequately capture the complexities of online interactions. Advanced methods, including attention mechanisms and graph neural networks, enhance accuracy but typically overlook user psychology and behavioral dynamics. Large language models (LLMs), with their human-like reasoning, offer new potential for simulating psychological aspects of information spread. We introduce an LLM-based simulation environment capturing agents' evolving attitudes, emotions, and responses. Initial experiments, however, revealed significant gaps between LLM-generated behaviors and authentic human dynamics, especially in stance detection and psychological realism. A detailed evaluation through Social Information Processing Theory identified major discrepancies in goal-setting and feedback evaluation, stemming from the lack of emotional processing in standard LLM training. To address these issues, we propose the Social Information Processing-based Chain of Thought (SIP-CoT) mechanism enhanced by emotion-guided memory. This method improves the interpretation of social cues, personalization of goals, and evaluation of feedback. Experimental results confirm that SIP-CoT-enhanced LLM agents more effectively process social information, demonstrating behaviors, attitudes, and emotions closer to real human interactions. In summary, this research highlights critical limitations in current LLM-based propagation simulations and demonstrates how integrating SIP-CoT and emotional memory significantly enhances the social intelligence and realism of LLM agents.

cs.AI↗

arXiv2025

Loki's Dance of Illusions: A Comprehensive Survey of Hallucination in Large Language Models

Edgar Allan Poe noted, "Truth often lurks in the shadow of error," highlighting the deep complexity intrinsic to the interplay between truth and falsehood, notably under conditions of cognitive and informational asymmetry. This dynamic is strikingly evident in large language models (LLMs). Despite their impressive linguistic generation capabilities, LLMs sometimes produce information that appears factually accurate but is, in reality, fabricated, an issue often referred to as 'hallucinations'. The prevalence of these hallucinations can mislead users, affecting their judgments and decisions. In sectors such as finance, law, and healthcare, such misinformation risks causing substantial economic losses, legal disputes, and health risks, with wide-ranging consequences.In our research, we have methodically categorized, analyzed the causes, detection methods, and solutions related to LLM hallucinations. Our efforts have particularly focused on understanding the roots of hallucinations and evaluating the efficacy of current strategies in revealing the underlying logic, thereby paving the way for the development of innovative and potent approaches. By examining why certain measures are effective against hallucinations, our study aims to foster a comprehensive approach to tackling this issue within the domain of LLMs.

cs.CL↗

arXiv2025

Lifelong Evolution: Collaborative Learning between Large and Small Language Models for Continuous Emergent Fake News Detection

The widespread dissemination of fake news on social media has significantly impacted society, resulting in serious consequences. Conventional deep learning methodologies employing small language models (SLMs) suffer from extensive supervised training requirements and difficulties adapting to evolving news environments due to data scarcity and distribution shifts. Large language models (LLMs), despite robust zero-shot capabilities, fall short in accurately detecting fake news owing to outdated knowledge and the absence of suitable demonstrations. In this paper, we propose a novel Continuous Collaborative Emergent Fake News Detection (C$^2$EFND) framework to address these challenges. The C$^2$EFND framework strategically leverages both LLMs' generalization power and SLMs' classification expertise via a multi-round collaborative learning framework. We further introduce a lifelong knowledge editing module based on a Mixture-of-Experts architecture to incrementally update LLMs and a replay-based continue learning method to ensure SLMs retain prior knowledge without retraining entirely. Extensive experiments on Pheme and Twitter16 datasets demonstrate that C$^2$EFND significantly outperforms existed methods, effectively improving detection accuracy and adaptability in continuous emergent fake news scenarios.

cs.CL↗

arXiv2025

Diffusion with a Linguistic Compass: Steering the Generation of Clinically Plausible Future sMRI Representations for Early MCI Conversion Prediction

Early prediction of Mild Cognitive Impairment (MCI) conversion is hampered by a trade-off between immediacy--making fast predictions from a single baseline sMRI--and accuracy--leveraging longitudinal scans to capture disease progression. We propose MCI-Diff, a diffusion-based framework that synthesizes clinically plausible future sMRI representations directly from baseline data, achieving both real-time risk assessment and high predictive performance. First, a multi-task sequence reconstruction strategy trains a shared denoising network on interpolation and extrapolation tasks to handle irregular follow-up sampling and learn robust latent trajectories. Second, an LLM-driven "linguistic compass" is introduced for clinical plausibility sampling: generated feature candidates are quantized, tokenized, and scored by a fine-tuned language model conditioned on expected structural biomarkers, guiding autoregressive generation toward realistic disease patterns. Experiments on ADNI and AIBL cohorts show that MCI-Diff outperforms state-of-the-art baselines, improving early conversion accuracy by 5-12%.

cs.LG↗

arXiv2025

MirrorShield: Towards Universal Defense Against Jailbreaks via Entropy-Guided Mirror Crafting

Defending large language models (LLMs) against jailbreak attacks is crucial for ensuring their safe deployment. Existing defense strategies typically rely on predefined static criteria to differentiate between harmful and benign prompts. However, such rigid rules fail to accommodate the inherent complexity and dynamic nature of real-world jailbreak attacks. In this paper, we focus on the novel challenge of universal defense against diverse jailbreaks. We propose a new concept ``mirror'', which is a dynamically generated prompt that reflects the syntactic structure of the input while ensuring semantic safety. The discrepancies between input prompts and their corresponding mirrors serve as guiding principles for defense. A novel defense model, MirrorShield, is further proposed to detect and calibrate risky inputs based on the crafted mirrors. Evaluated on multiple benchmark datasets and compared against ten state-of-the-art attack methods, MirrorShield demonstrates superior defense performance and promising generalization capabilities.

cs.CR↗

arXiv2025

Collaborative Evolution: Multi-Round Learning Between Large and Small Language Models for Emergent Fake News Detection

The proliferation of fake news on social media platforms has exerted a substantial influence on society, leading to discernible impacts and deleterious consequences. Conventional deep learning methodologies employing small language models (SLMs) suffer from the necessity for extensive supervised training and the challenge of adapting to rapidly evolving circumstances. Large language models (LLMs), despite their robust zero-shot capabilities, have fallen short in effectively identifying fake news due to a lack of pertinent demonstrations and the dynamic nature of knowledge. In this paper, a novel framework Multi-Round Collaboration Detection (MRCD) is proposed to address these aforementioned limitations. The MRCD framework is capable of enjoying the merits from both LLMs and SLMs by integrating their generalization abilities and specialized functionalities, respectively. Our approach features a two-stage retrieval module that selects relevant and up-to-date demonstrations and knowledge, enhancing in-context learning for better detection of emerging news events. We further design a multi-round learning framework to ensure more reliable detection results. Our framework MRCD achieves SOTA results on two real-world datasets Pheme and Twitter16, with accuracy improvements of 7.4\% and 12.8\% compared to using only SLMs, which effectively addresses the limitations of current models and improves the detection of emergent fake news.

cs.CL↗

arXiv2024

HNCSE: Advancing Sentence Embeddings via Hybrid Contrastive Learning with Hard Negatives

Unsupervised sentence representation learning remains a critical challenge in modern natural language processing (NLP) research. Recently, contrastive learning techniques have achieved significant success in addressing this issue by effectively capturing textual semantics. Many such approaches prioritize the optimization using negative samples. In fields such as computer vision, hard negative samples (samples that are close to the decision boundary and thus more difficult to distinguish) have been shown to enhance representation learning. However, adapting hard negatives to contrastive sentence learning is complex due to the intricate syntactic and semantic details of text. To address this problem, we propose HNCSE, a novel contrastive learning framework that extends the leading SimCSE approach. The hallmark of HNCSE is its innovative use of hard negative samples to enhance the learning of both positive and negative samples, thereby achieving a deeper semantic understanding. Empirical tests on semantic textual similarity and transfer task datasets validate the superiority of HNCSE.

cs.CL↗

arXiv2024

FineFake: A Knowledge-Enriched Dataset for Fine-Grained Multi-Domain Fake News Detection

Existing benchmarks for fake news detection have significantly contributed to the advancement of models in assessing the authenticity of news content. However, these benchmarks typically focus solely on news pertaining to a single semantic topic or originating from a single platform, thereby failing to capture the diversity of multi-domain news in real scenarios. In order to understand fake news across various domains, the external knowledge and fine-grained annotations are indispensable to provide precise evidence and uncover the diverse underlying strategies for fabrication, which are also ignored by existing benchmarks. To address this gap, we introduce a novel multi-domain knowledge-enhanced benchmark with fine-grained annotations, named \textbf{FineFake}. FineFake encompasses 16,909 data samples spanning six semantic topics and eight platforms. Each news item is enriched with multi-modal content, potential social context, semi-manually verified common knowledge, and fine-grained annotations that surpass conventional binary labels. Furthermore, we formulate three challenging tasks based on FineFake and propose a knowledge-enhanced domain adaptation network. Extensive experiments are conducted on FineFake under various scenarios, providing accurate and reliable benchmarks for future endeavors. The entire FineFake project is publicly accessible as an open-source repository at \url{https://github.com/Accuser907/FineFake}.

cs.CL↗

arXiv2024

Beyond Entity Alignment: Towards Complete Knowledge Graph Alignment via Entity-Relation Synergy

Knowledge Graph Alignment (KGA) aims to integrate knowledge from multiple sources to address the limitations of individual Knowledge Graphs (KGs) in terms of coverage and depth. However, current KGA models fall short in achieving a ``complete'' knowledge graph alignment. Existing models primarily emphasize the linkage of cross-graph entities but overlook aligning relations across KGs, thereby providing only a partial solution to KGA. The semantic correlations embedded in relations are largely overlooked, potentially restricting a comprehensive understanding of cross-KG signals. In this paper, we propose to conceptualize relation alignment as an independent task and conduct KGA by decomposing it into two distinct but highly correlated sub-tasks: entity alignment and relation alignment. To capture the mutually reinforcing correlations between these objectives, we propose a novel Expectation-Maximization-based model, EREM, which iteratively optimizes both sub-tasks. Experimental results on real-world datasets demonstrate that EREM consistently outperforms state-of-the-art models in both entity alignment and relation alignment tasks.

cs.CL↗