arXiv subjects

Rui Pu

Publications and source records attributed to Rui Pu.

9 recordsLinked to original sources

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

Localizing Individual Exciton on a Quantum Hall Antidot

Quantum Hall systems host quasiparticles demonstrating correlated electron physics and non-trivial quantum statistics. Excitonic phases, archetypical for interaction effect, have attracted significant interest in recent years in double-layer quantum Hall systems where spatially separated electrons and holes form bosonic condensate through Coulomb interaction. Here, employing the approach of quantum Hall antidot with two spatially separated edge channels, we demonstrate a new type of quantum Hall quasiparticle exciton which represents a quantum-coherent bound state of an electron and a hole situated on their corresponding edges coupled through intralayer tunneling and Coulomb interaction. Quantum-coherent dynamics of the exciton is reflected in the observed evolution of the position and magnitude of the antidot conductance peaks around the electron-hole resonance. The quantum Hall antidot setup allows localization and electrical tuning of individual quantum Hall excitons. Quantum superposition of vacuum- and electron-hole pairing states is observed through the gate-dependent tunneling conductance of the antidot. Modeling the electron-hole pair as a coupled two-level system, semi-quantitative understanding of experimental observations is achieved. This work opens avenues for creating quantum systems of multiple quantum Hall quasiparticles.

cond-mat.mes-hall↗

arXiv2025

From "Aha Moments" to Controllable Thinking: Toward Meta-Cognitive Reasoning in Large Reasoning Models via Decoupled Reasoning and Control

Large Reasoning Models (LRMs) can exhibit step-by-step reasoning, reflection, and backtracking, but these behaviors are often unregulated, leading to overthinking. As a result, LRMs continue generating redundant reasoning even after reaching high-confidence conclusions. This increases inference cost and latency, limiting practical deployment. The root cause is the absence of an intrinsic mechanism to monitor the reasoning state and decide when to continue, backtrack, or stop. We propose MERA, a meta-cognitive reasoning framework that decouples reasoning from control to enable independent optimization of control strategies. MERA constructs high-quality reasoning-control supervision data via a takeover-based pipeline, and transforms long-horizon traces into structured reasoning-control alternating sequences for training. The model is trained with supervised fine-tuning to internalize the structured separation, and further optimized with Control-Segment Policy Optimization (CSPO), which combines segment-wise GRPO with control masking to focus learning on control segments. Experiments across reasoning benchmarks show that MERA improves both efficiency and accuracy.

cs.AI↗

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

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↗

arXiv2024

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↗

arXiv2023

Nano-Imaging of Landau-Phonon Polaritons in Dirac Heterostructures

Polaritons are light-matter quasiparticles that govern the optical response of quantum materials and enable their nanophotonic applications. We have studied a new type of polaritons arising in magnetized graphene encapsulated in hexagonal boron nitride (hBN). These polaritons stem from hybridization of Dirac magnetoexciton modes of graphene with waveguide phonon modes of hBN crystals. We refer to these quasiparticles as the Landau-phonon polaritons (LPPs). Using infrared magneto nanoscopy, we imaged LPPs and controlled their real-space propagation by varying the magnetic field. These LLPs have large in-plane momenta and are not bound by the conventional optical selection rules, granting us access to the "forbidden" inter-Landau level transitions (ILTs). We observed avoided crossings in the LPP dispersion - a hallmark of the strong coupling regime - occurring when the magnetoexciton and hBN phonon frequencies matched. Our LPP-based nanoscopy also enabled us to resolve two fundamental many-body effects: the graphene Fermi velocity renormalization and ILT-dependent magnetoexciton binding energies. These results indicate that magnetic-field-tuned Dirac heterostructures are promising platforms for precise nanoscale control and sensing of light-matter interaction.

physics.optics↗

arXiv2020

A Full Second-Order Analysis of the Widely Linear MVDR Beamformer for Noncircular Signals

A full performance analysis of the widely linear (WL) minimum variance distortionless response (MVDR) beamformer is introduced. While the WL MVDR is known to outperform its strictly linear counterpart, the Capon beamformer, for noncircular complex signals, the existing approaches provide limited physical insights, since they explicitly or implicitly omit the complementary second-order (SO) statistics of the output interferences and noise (IN). To this end, we exploit the full SO statistics of the output IN to introduce a full SO performance analysis framework for the WL MVDR beamformer. This makes it possible to separate the overall signal-to-interference plus noise ratio (SINR) gain of the WL MVDR beamformer w.r.t. the Capon one into the individual contributions along the in-phase (I) and quadrature (Q) channels. Next, by considering the reception of the unknown signal of interest (SOI) corrupted by an arbitrary number of orthogonal noncircular interferences, we further unveil the distribution of SINR gains in both the I and Q channels, and show that in almost all the spatial cases, these performance advantages are more pronounced when the SO noncircularity rate of the interferences increases. Illustrative numerical simulations are provided to support the theoretical results.

cs.IT↗