arXiv subjects

Qingyu Meng

Publications and source records attributed to Qingyu Meng.

8 recordsLinked to original sources

arXiv2026

SEAL: Reinforcing Global Safety in Mixture-of-Experts through Shared Expert ALignment

Mixture-of-Experts (MoE) is a scaling architecture for large language models that activates only a small subset of expert modules per token, enabling massive parameter growth with nearly constant computation. Recent Hybrid MoE architecture adds \textit{shared experts} to capture consistently useful representations, further improving stability and generalization. MoE now powers many flagship open-source and commercial models, yet remains vulnerable to adversarial attacks. Specifically, sparse routing introduces a structural vulnerability: MoE safety hinges on which experts are activated, and adversaries can subvert this selection through jailbreak prompts, malicious fine-tuning, and weight-level pruning of safety-critical neurons. Existing defenses primarily focus on hardening the router, but an adversary may still manipulate or bypass the routing trajectory due to the routing process's nondeterministic nature, thereby collapsing the defense. To cope with this problem, we first identify theoretically and empirically that shared expert, an always-activated component containing a small proportion of safety-critical neurons, can overcome the uncertainty of sparsely activated routing path and serve as a router-independent anchor to enhance global safety alignment. Based on this insight, we propose SEAL, a training-time parameter-efficient defense that produces a plug-and-play adapter attached to shared expert, and SEAL++, a variant that adds an orthogonal constraint preserving pre-existing safety subspaces during training. We evaluate SEAL and SEAL++ across six attack scenarios that combine three adversarial inputs (harmful prompting, jailbreak, malicious fine-tuning) with and without neuron pruning. SEAL reduces attack success rate (ASR) by up to 60\%, at a capability cost of at most 1.4\% on a five-benchmark average. Additionally, SEAL can seamlessly integrate with router-level ......

cs.LG↗

arXiv2026

SciText2Eq: Assessing LLMs for Explainable Equation Generation for Scientific Creativity

This work investigates the ability of large language models (LLMs) to generate mathematical equations from scientific texts. Prior work faces challenges in unstructured grounding, multi-equation dependency, and humanaligned evaluation. To this end, we construct a dataset of AI research papers, pairing contextual passages with ground-truth equations and variable descriptions. We develop an explainable equation generation workflow and evaluate it across diverse open- and closed-source LLM backbones. We introduce an evaluation protocol combining automatic metrics, LLM-based rubrics, and human judgments to assess accuracy, explainability, and human-LLM alignment. Results indicate that LLMs perform moderately on lexical- and syntactic-based similarity, while struggling with semantic accuracy. Comparisons between LLM-based evaluations and human judgments reveal limited alignment, highlighting challenges in using LLMs to assess equation quality. These findings offer insights for improving equation generation models and developing more reliable evaluation methods for scientific text. We provide code and data for reproducibility.

cs.AI↗

arXiv2026

Trust Stack for Mental Health AI: A Survey of Calibration across Human, Interaction, and AI Layers

Language-based AI is increasingly deployed for mental health support, yet trust is evaluated in interdisciplinary but operationally misaligned ways: NLP and AI work measures robustness, safety, privacy, and explanations, while psychotherapy, HCI, and regulatory work emphasize therapeutic fidelity, lived experience, empathy, and reliance. Empathetic chatbots can elicit strong user trust without commensurate safety, while safer systems are under-trusted when their boundaries are opaque, a calibration gap no single community owns. Through a structured scoping synthesis of 61 papers, we survey this landscape into a three-layer framework separating (L1) human-oriented trust, (L2) interaction-oriented trustworthiness, and (L3) AI-oriented trustworthiness, and map five stakeholder perspectives onto these layers. We outline a research agenda for building socio-technically aligned trustworthy AI for mental health support, highlighting that the central objective should shift from maximizing perceived trust to calibrating human trust to demonstrated interaction- and AI-level trustworthiness.

cs.CL↗

arXiv2026

StoryMI: Steerable Multi-Agent Therapeutic Dialogue Generation

Large language models (LLMs) can generate fluent dialogue, but prior works lack situational grounding, dynamic strategy control, and evaluation aligned with clinical standards in motivational interviewing (MI). We introduce StoryMI, a multi-LLM agent framework for controllable MI dialogue generation, where questionnaire-based client profiles are expanded into situational stories that provide narrative context for the dialogue. Therapist and client agents generate MI-coded utterances guided by MI codes selected by the interaction agent, while an interaction agent dynamically coordinates exchanges to control MI strategies during a multi-turn conversation. We propose a two-level evaluation protocol: lexical metrics and MI-specific measures of macro-level counseling strategies, alongside LLM-as-judge and human expert assessments. We construct a dataset of 6K simulated MI dialogues grounded in 1K questionnaire-story pairs, covering 12 MI codes and 13 symptom domains, and benchmark six open- and closed-source LLMs. Our results show that situational grounding and macro-level control can improve MI adherence and clinical plausibility, demonstrating the effectiveness of a structured multi-agent workflow for psychotherapy dialogue generation. We provide code and data for reproducibility.

cs.CL↗

arXiv2026

Geometric Preconditioning and Curriculum Optimization for Trainable Variational Quantum Regression

Variational quantum circuits are increasingly studied as continuous-function approximators, but quantum regression remains difficult to train when global losses, finite-shot stochasticity, and circuit-depth growth combine to produce weak or ill-conditioned gradient signals. We study this trainability problem in a controlled hybrid quantum--classical regression design. The central ingredient is a capacity-controlled classical embedding that acts as a learnable geometric preconditioner: it reshapes the input distribution seen by a data-reuploading variational circuit while preserving a low-dimensional quantum bottleneck. We pair this representation design with a curriculum protocol that grows circuit depth progressively and switches from SPSA-based stochastic exploration to Adam-based analytic-gradient fine-tuning. We formalize the mechanism through a local quantum-tangent contraction statement: in the linearized quantum-parameter dynamics, the embedding changes the empirical Gram matrix that controls residual contraction and one-step loss decrease. Across finite-size statevector audits on PDE-informed regression benchmarks and small-data tabular tasks, the Hybrid QNN lowers error relative to Pure QNN baselines under matched quantum-model budgets. Strong classical references remain competitive, and in several cases are better in absolute error; the evidence therefore supports a trainability claim for the hybrid QNN design rather than a claim of classical or hardware quantum advantage.

cs.LG↗

arXiv2026

MAFS: Multi-head Attention Feature Selection for High-Dimensional Data via Deep Fusion of Filter Methods

Feature selection is essential for high-dimensional biomedical data, enabling stronger predictive performance, reduced computational cost, and improved interpretability in precision medicine applications. Existing approaches face notable challenges. Filter methods are highly scalable but cannot capture complex relationships or eliminate redundancy. Deep learning-based approaches can model nonlinear patterns but often lack stability, interpretability, and efficiency at scale. Single-head attention improves interpretability but is limited in capturing multi-level dependencies and remains sensitive to initialization, reducing reproducibility. Most existing methods rarely combine statistical interpretability with the representational power of deep learning, particularly in ultra-high-dimensional settings. Here, we introduce MAFS (Multi-head Attention-based Feature Selection), a hybrid framework that integrates statistical priors with deep learning capabilities. MAFS begins with filter-based priors for stable initialization and guide learning. It then uses multi-head attention to examine features from multiple perspectives in parallel, capturing complex nonlinear relationships and interactions. Finally, a reordering module consolidates outputs across attention heads, resolving conflicts and minimizing information loss to generate robust and consistent feature rankings. This design combines statistical guidance with deep modeling capacity, yielding interpretable importance scores while maximizing retention of informative signals. Across simulated and real-world datasets, including cancer gene expression and Alzheimer's disease data, MAFS consistently achieves superior coverage and stability compared with existing filter-based and deep learning-based alternatives, offering a scalable, interpretable, and robust solution for feature selection in high-dimensional biomedical data.

cs.LG↗

arXiv2025

Efficient Mathematical Reasoning Models via Dynamic Pruning and Knowledge Distillation

With the rapid development of deep learning, large language models have shown strong capabilities in complex reasoning tasks such as mathematical equation solving. However, their substantial computational and storage costs hinder practical deployment. This paper proposes a lightweight optimization method that integrates dynamic attention head pruning with knowledge distillation. The approach dynamically evaluates the importance of each attention head in the multi-head attention mechanism using a combination of weight norms and entropy, and prunes redundant heads in real time to reduce computational overhead. To mitigate performance degradation, knowledge distillation transfers information from the original model to the pruned student, enabling the smaller model to preserve reasoning ability. Experiments conducted on both Math23k and ASDiv-A verify the effectiveness of the proposed method. For example, on Math23k with a 30% pruning ratio, parameters are reduced by 18.7%, inference speed is improved by 27.5%, FLOPs are reduced by 19.3%, and accuracy drops only 0.7% (from 84.4% to 83.7%). These results demonstrate that the method achieves substantial efficiency gains while maintaining strong reasoning performance, providing a practical solution for efficient deployment of large language models in mathematical reasoning tasks.

cs.LG↗

arXiv2022

Wind power ramp prediction algorithm based on wavelet deep belief network

The wind power ramp events threaten the power grid safety significantly. To improve the ramp prediction accuracy, a hybrid wavelet deep belief network algorithm with adaptive feature selection (WDBNAFS) is proposed. First, the wind power characteristic is analyzed. Then, wavelet decomposition is addressed to the time series, and an adaptive feature selection algorithm is proposed to select the inputs of the prediction model. Finally, a deep belief network is employed to predict the wind power ramp event, and the proposed WDBNAFS was testified with the experiments based on the practical data. The simulation results demonstrate that the prediction accuracy of the proposed algorithm is more than 90%.

eess.SY↗