arXiv subjects

Zhigang Zeng

Publications and source records attributed to Zhigang Zeng.

At least 19 recordsLinked to original sources

arXiv2026

Synergistic Perception-Reasoning Governance: Grounding Medical MLLMs with Verifiable Anatomical Evidence

Multimodal large language models (MLLMs) show strong promise for clinical VQA and radiology report generation, yet inference-time hallucinations still undermine trustworthy use: models can produce fluent conclusions that conflict with imaging evidence. Existing mitigation strategies typically rely on additional training, external retrieval/knowledge bases, or multi-stage post-hoc verification, which increases cost and pipeline complexity and often generalizes poorly across models and tasks.To address this, we propose a holistic, training-free evidence-injection framework that systematically mitigates hallucinations through dual-side evidence injection. By leveraging ROI priors acquired using MedSAM in our implementation, we recalibrate the visual perception trajectory via ROI-guided activation modulation while anchoring the textual reasoning trajectory by mapping anatomical coordinates into discrete semantic tokens as verifiable external memory. Then we introduce a task-aware dynamic router to select modality-specific interventions based on task semantics, balancing perceptual grounding and linguistic fluency. We conduct systematic evaluations on 2 tasks and 5 datasets using \texttt{LLaVA-1.5-7B}, \texttt{LLaVA-Med-1.5-7B}, \texttt{Qwen3-VL-8B/32B}, and \texttt{InternVL-3.5-8B/38B}. Controlled ablations and visualizations further validate the framework, which consistently outperforms baselines across medical benchmarks, improving close-ended accuracy by up to $\sim\mathbf{6}\%\uparrow$ and reducing open-ended hallucinations by $\sim\mathbf{35}\%\downarrow$. The code has been made available on GitHub: \href{https://github.com/Henry991115/SPRG}{\textcolor{blue}{https://github.com/Henry991115/SPRG}}.

cs.CV↗

arXiv2026

A Survey of Large Audio Language Models: Generalization, Trustworthiness, and Outlook

Advances in Large Language Models (LLMs) have paved the way for Multimodal Large Language Models (MLLMs). Among these, Large Audio Language Models (LALMs) are essential for realizing universal auditory intelligence. Despite their remarkable performance, the escalation of LALMs' capabilities has significantly outpaced the development of systemic frameworks to ensure their trustworthiness. This survey provides a comprehensive investigation into the endogenous mechanisms of LALMs, detailing the architectural innovations and alignment algorithms that facilitate emergent reasoning. Specifically, we analyze how the transition to unified end-to-end frameworks and the integration of continuous acoustic signals expand the attack surface. To rigorously evaluate the risks within these paradigms, we establish a comprehensive taxonomy of trustworthiness, categorizing critical vulnerabilities such as cross-modal jailbreaking, latent acoustic backdoors, and biometric privacy leakage. We review the state-of-the-art LALMs through six analytical pillars: hallucination, robustness, safety, privacy, fairness, and authentication. The pronounced imbalance between a mature offensive landscape and underdeveloped defenses highlights persistent trustworthiness gaps and multidimensional risks in audio-centric intelligence. Finally, we propose a roadmap advocating for ``Defense-in-Depth'' architectures, causal auditory world modeling, and intrinsic representation engineering to support the development of more reliable and trustworthy audio intelligence. Our project has been uploaded to GitHub https://github.com/Kwwwww74/Awesome-Trustworthy-AudioLLMs.

cs.SD↗

arXiv2026

Explaining and Breaking the Safety-Helpfulness Ceiling via Preference Dimensional Expansion

In the realm of multi-objective alignment for large language models, balancing disparate human preferences often manifests as a zero-sum conflict. Specifically, the intrinsic tension between competing goals dictates that aggressively optimizing for one metric (e.g., helpfulness) frequently incurs a substantial penalty on another (e.g., harmlessness). While prior work mainly focuses on data selection, parameter merging, or algorithmic balancing during training, these approaches merely force compromises between divergent preferences along a fixed Pareto frontier, failing to fundamentally resolve the inherent trade-off. In this work, we approach this problem from a novel perspective of multi-dimensional rewards. By scaling up the model's rollouts and analyzing the outputs across different reward dimensions, we arrive at a critical conclusion: the conflict among multiple objectives stems from the fact that the prompt itself inherently restricts the achievable multi-dimensional rewards. Based on this core observation, we propose MORA: Multi-Objective Reward Assimilation. Specifically, MORA isolates single-reward prompts through pre-sampling and expands their reward diversity by rewriting the original questions to incorporate multi-dimensional intents. Extensive experiments demonstrate that: (1) in sequential alignment, MORA achieves single-preference improvements ranging from 5% to 12.4%, with exceptional gains in harmlessness, after multiple-preference alignment across helpful, harmless, and truthful dimensions. (2) In simultaneous alignment, MORA achieves an average overall reward improvement of 4.6%. Our codes are available at https://github.com/Shiying-Huang/MORA-MPA.

cs.AI↗

arXiv2026

SaFeR-Steer: Evolving Multi-Turn MLLMs via Synthetic Bootstrapping and Feedback Dynamics

MLLMs are increasingly deployed in multi-turn settings, where attackers can escalate unsafe intent through the evolving visual-text history and exploit long-context safety decay. Yet safety alignment is still dominated by single-turn data and fixed-template dialogues, leaving a mismatch between training and deployment. To bridge this gap, we propose SaFeR-Steer, a progressive multi-turn alignment framework that combines staged synthetic bootstrapping with tutor-in-the-loop GRPO to train a single student under adaptive, on-policy attacks. We also introduce Trajectory-Consistent Summative Reward (TCSR), which aggregates the historical minimum and average of turn rewards so that any low-quality turn affects the trajectory-level return. I. Dataset. We release STEER, a multi-turn multimodal safety dataset with STEER-SFT (12,934), STEER-RL (2,000), and STEER-Bench (3,227) dialogues spanning 2-10 turns. II. Experiment. Starting from Qwen2.5-VL-3B/7B, SaFeR-Steer substantially improves Safety/Helpfulness on both single-turn (48.30/45.86 $\rightarrow$ 81.84/70.77 for 3B; 56.21/60.32 $\rightarrow$ 87.89/77.40 for 7B) and multi-turn benchmarks (12.55/27.13 $\rightarrow$ 55.58/70.27 for 3B; 24.66/46.48 $\rightarrow$ 64.89/72.35 for 7B), shifting failures to later turns and yielding robustness beyond scaling alone. Code is available at https://anonymous.4open.science/r/SaFeR-Steer

cs.LG↗

arXiv2026

PathMem: Toward Cognition-Aligned Memory Transformation for Pathology MLLMs

Computational pathology demands both visual pattern recognition and dynamic integration of structured domain knowledge, including taxonomy, grading criteria, and clinical evidence. In practice, diagnostic reasoning requires linking morphological evidence with formal diagnostic and grading criteria. Although multimodal large language models (MLLMs) demonstrate strong vision language reasoning capabilities, they lack explicit mechanisms for structured knowledge integration and interpretable memory control. As a result, existing models struggle to consistently incorporate pathology-specific diagnostic standards during reasoning. Inspired by the hierarchical memory process of human pathologists, we propose PathMem, a memory-centric multimodal framework for pathology MLLMs. PathMem organizes structured pathology knowledge as a long-term memory (LTM) and introduces a Memory Transformer that models the dynamic transition from LTM to working memory (WM) through multimodal memory activation and context-aware knowledge grounding, enabling context-aware memory refinement for downstream reasoning. PathMem achieves SOTA performance across benchmarks, improving WSI-Bench report generation (12.8% WSI-Precision, 10.1% WSI-Relevance) and open-ended diagnosis by 9.7% and 8.9% over prior WSI-based models.

cs.AI↗

arXiv2026

Hit-RAG: Learning to Reason with Long Contexts via Preference Alignment

Despite the promise of Retrieval-Augmented Generation in grounding Multimodal Large Language Models with external knowledge, the transition to extensive contexts often leads to significant attention dilution and reasoning hallucinations. The surge in information density causes critical evidence to be submerged by voluminous noise, which complicates the discernment of relevant fragments within a dense input. In this paper, we propose \textbf{Hit-RAG}, a multi-stage preference alignment framework designed to resolve these cognitive bottlenecks through a progressive optimization pipeline. Our approach systematically refines the utilization of external evidence via three distinct stages. First, Supervised Fine-tuning establishes baseline context awareness to minimize information neglect. Next, Discriminative Preference Alignment enhances robustness against misleading distractors. Finally, Group-Relative Policy Optimization stabilizes logical synthesis to prevent reasoning collapse. Extensive evaluations on eight benchmarks demonstrate that Hit-RAG consistently yields substantial performance gains, enabling models to bridge the gap between context acquisition and accurate reasoning while surpassing much larger counterparts in long-context scenarios.

cs.CL↗

arXiv2026

SaFeR-ToolKit: Structured Reasoning via Virtual Tool Calling for Multimodal Safety

Vision-language models remain susceptible to multimodal jailbreaks and over-refusal because safety hinges on both visual evidence and user intent, while many alignment pipelines supervise only the final response. To address this, we present SaFeR-ToolKit, which formalizes safety decision-making as a checkable protocol. Concretely, a planner specifies a persona, a Perception $\to$ Reasoning $\to$ Decision tool set, and a constrained transition graph, while a responder outputs a typed key-value tool trace before the final answer. To make the protocol reliably followed in practice, we train a single policy with a three-stage curriculum (SFT $\to$ DPO $\to$ GRPO), where GRPO directly supervises tool usage beyond answer-level feedback. Our contributions are two-fold: I. Dataset. The first tool-based safety reasoning dataset, comprising 31,654 examples (SFT 6k, DPO 18.6k, GRPO 6k) plus 1k held-out evaluation. II. Experiments. On Qwen2.5-VL, SaFeR-ToolKit significantly improves Safety/Helpfulness/Reasoning Rigor on 3B (29.39/45.04/4.98 $\to$ 84.40/71.13/78.87) and 7B (53.21/52.92/19.26 $\to$ 86.34/80.79/85.34), while preserving general capabilities (3B: 58.67 $\to$ 59.21; 7B: 66.39 $\to$ 66.81). Codes are available at https://github.com/Duebassx/SaFeR_ToolKit.

cs.LG↗

arXiv2026

Diagnosing Knowledge Conflict in Multimodal Long-Chain Reasoning

Multimodal large language models (MLLMs) in long chain-of-thought reasoning often fail when different knowledge sources provide conflicting signals. We formalize these failures under a unified notion of knowledge conflict, distinguishing input-level objective conflict from process-level effective conflict. Through probing internal representations, we reveal that: (I) Linear Separability: different conflict types are explicitly encoded as linearly separable features rather than entangled; (II) Depth Localization: conflict signals concentrate in mid-to-late layers, indicating a distinct processing stage for conflict encoding; (III) Hierarchical Consistency: aggregating noisy token-level signals along trajectories robustly recovers input-level conflict types; and (IV) Directional Asymmetry: reinforcing the model's implicit source preference under conflict is far easier than enforcing the opposite source. Our findings provide a mechanism-level view of multimodal reasoning under knowledge conflict and enable principled diagnosis and control of long-CoT failures.

cs.AI↗

arXiv2026

StackingNet: Collective Inference Across Independent AI Foundation Models

Artificial intelligence built on large foundation models has transformed language understanding, computer vision, and reasoning, yet these systems remain isolated and cannot readily share their capabilities. Coordinating the complementary strengths of independently developed, black-box foundation models is essential for trustworthy intelligent systems, yet no established method exists. Here we show that such coordination can be achieved through a meta-ensemble framework termed StackingNet, which aggregates the output predictions of independent models at inference. StackingNet improves accuracy, reduces individual-model error and group-wise disparities, ranks model reliability, and identifies or prunes models that degrade performance, all without access to internal parameters or training data. Across language comprehension, visual attribute estimation, and academic paper rating, it consistently outperforms individual models and classic ensembles, with gains that persist when the base models are uniformly strong. These gains stem from variance reduction and consensus alignment among independent models rather than from any emergent group cognition, and they widen as the model pool grows more diverse. By turning model diversity from a source of inconsistency into a resource for cooperation, StackingNet offers a practical path toward coordinated artificial intelligence, where progress emerges not only from larger single models but from principled cooperation among many specialized ones.

cs.AI↗

arXiv2026

CSSBench: Evaluating the Safety of Lightweight LLMs against Chinese-Specific Adversarial Patterns

Large language models (LLMs) are increasingly deployed in cost-sensitive and on-device scenarios, and safety guardrails have advanced mainly in English. However, real-world Chinese malicious queries typically conceal intent via homophones, pinyin, symbol-based splitting, and other Chinese-specific patterns. These Chinese-specific adversarial patterns create the safety evaluation gap that is not well captured by existing benchmarks focused on English. This gap is particularly concerning for lightweight models, which may be more vulnerable to such specific adversarial perturbations. To bridge this gap, we introduce the Chinese-Specific Safety Benchmark (CSSBench) that emphasizes these adversarial patterns and evaluates the safety of lightweight LLMs in Chinese. Our benchmark covers six domains that are common in real Chinese scenarios, including illegal activities and compliance, privacy leakage, health and medical misinformation, fraud and hate, adult content, and public and political safety, and organizes queries into multiple task types. We evaluate a set of popular lightweight LLMs and measure over-refusal behavior to assess safety-induced performance degradation. Our results show that the Chinese-specific adversarial pattern is a critical challenge for lightweight LLMs. This benchmark offers a comprehensive evaluation of LLM safety in Chinese, assisting robust deployments in practice.

cs.CL↗

arXiv2025

Checkerboard-type Zhang-Rice States in Overdoped Cuprate Superconductors

Cuprate superconductors remain central to condensed matter physics due to their technological relevance and unconventional, incompletely understood electronic behavior. While the canonical phase diagram and low-energy models have been shaped largely by studies of underdoped and moderately doped cuprates, the overdoped regime has received comparatively limited attention.Here, we track the evolution of the electronic structure from optimal to heavy overdoping in La2-xSrxCuO4(LSCO) using broadband optical spectroscopy across x=0.15-0.60. The measured spectral changes--including the redistribution of Zhang-Rice-related spectral weigh--are in qualitative agreement with determinant quantum Monte Carlo simulations of the three-orbital Emery model, which together indicate a pronounced reconstruction of the electronic structure beyond hole concentrations x>0.2. Guided by these observations, we propose a spontaneous checkerboard-type Zhang-Rice electronic configuration that captures the coexistence of itinerant and localized carriers characteristic of the heavily overdoped state. Our results refine the doping-dependent Zhang-Rice-based framework for cuprates, illuminate how correlations persist deep into the overdoped regime, and provide new constraints on microscopic mechanisms of high-temperature superconductivity, with broader implications for correlated transition-metal oxides.

cond-mat.supr-con↗

arXiv2025

HD3C: Efficient Medical Data Classification for Edge Devices

Efficient medical data classification is essential for modern disease screening, particularly in resource-constrained environments where power budgets and computing capabilities are limited. We present HD3C, a lightweight classification framework designed for low-power edge devices. HD3C encodes data into high-dimensional hypervectors, aggregates them into multiple cluster prototypes, and performs classification through similarity search in hyperspace. We evaluate HD3C across three medical classification tasks; on heart sound classification, HD3C is 350x more energy-efficient than Bayesian ResNet with less than 1% difference in accuracy. Moreover, HD3C demonstrates exceptional robustness to noise, limited training data, and hardware error, supported by both theoretical analysis and empirical results, highlighting its potential for reliable deployment in real-world settings.

cs.LG↗

arXiv2025

Breaking Free from MMI: A New Frontier in Rationalization by Probing Input Utilization

Extracting a small subset of crucial rationales from the full input is a key problem in explainability research. The most widely used fundamental criterion for rationale extraction is the maximum mutual information (MMI) criterion. In this paper, we first demonstrate that MMI suffers from diminishing marginal returns. Once part of the rationale has been identified, finding the remaining portions contributes only marginally to increasing the mutual information, making it difficult to use MMI to locate the rest. In contrast to MMI that aims to reproduce the prediction, we seek to identify the parts of the input that the network can actually utilize. This is achieved by comparing how different rationale candidates match the capability space of the weight matrix. The weight matrix of a neural network is typically low-rank, meaning that the linear combinations of its column vectors can only cover part of the directions in a high-dimensional space (high-dimension: the dimensions of an input vector). If an input is fully utilized by the network, {it generally matches these directions (e.g., a portion of a hypersphere), resulting in a representation with a high norm. Conversely, if an input primarily falls outside (orthogonal to) these directions}, its representation norm will approach zero, behaving like noise that the network cannot effectively utilize. Building on this, we propose using the norms of rationale candidates as an alternative objective to MMI. Through experiments on four text classification datasets and one graph classification dataset using three network architectures (GRUs, BERT, and GCN), we show that our method outperforms MMI and its improved variants in identifying better rationales. We also compare our method with a representative LLM (llama-3.1-8b-instruct) and find that our simple method gets comparable results to it and can sometimes even outperform it.

cs.AI↗

arXiv2024

Multi-scale Cascaded Foundation Model for Whole-body Organs-at-risk Segmentation

Accurate segmentation of organs-at-risk (OARs) is vital for safe and precise radiotherapy and surgery. Most existing studies segment only a limited set of organs or regions, lacking a systematic treatment of OARs segmentation. We present a Multi-scale Cascaded Fusion Network (MCFNet) that aggregates features across multiple scales and resolutions. MCFNet consists of a Sharp Extraction Backbone for the downsampling path and a Flexible Connection Backbone for skip-connection fusion, strengthening representation learning in both stages. This design improves boundary localization and preserves fine structures while maintaining computational efficiency, enabling reliable performance even on low-resolution inputs. Experiments on an NVIDIA A6000 GPU using 36,131 image-mask pairs from 671 patients across 10 datasets show consistent robustness and strong cross-dataset generalization. An adaptive loss-aggregation strategy further stabilizes optimization and yields additional gains in accuracy and training efficiency. Through extensive validation, MCFNet outperforms existing methods, excelling in organ segmentation and providing reliable image-guided support for computer-aided diagnosis. Our solution aims to improve the precision and safety of radiotherapy and surgery while supporting personalized treatment, advancing modern medical technology. The code has been made available on GitHub: https://github.com/Henry991115/MCFNet.

eess.IV↗

arXiv2024

Large, Complex, and Realistic Safety Clothing and Helmet Detection: Dataset and Method

Detecting safety clothing and helmets is paramount for ensuring the safety of construction workers. However, the development of deep learning models in this domain has been impeded by the scarcity of high-quality datasets. In this study, we construct a large, complex, and realistic safety clothing and helmet detection (SFCHD) dataset. SFCHD is derived from two authentic chemical plants, comprising 12,373 images, 7 categories, and 50,552 annotations. We partition the SFCHD dataset into training and testing sets with a ratio of 4:1 and validate its utility by applying several classic object detection algorithms. Furthermore, drawing inspiration from spatial and channel attention mechanisms, we design a spatial and channel attention-based low-light enhancement (SCALE) module. SCALE is a plug-and-play component with a high degree of flexibility. Extensive evaluations of the SCALE module on both the ExDark and SFCHD datasets have empirically demonstrated its efficacy in enhancing the performance of detectors under low-light conditions. The dataset and code are publicly available at https://github.com/lijfrank-open/SFCHD-SCALE.

cs.CV↗

arXiv2024

Tracing Intricate Cues in Dialogue: Joint Graph Structure and Sentiment Dynamics for Multimodal Emotion Recognition

Multimodal emotion recognition in conversation (MERC) has garnered substantial research attention recently. Existing MERC methods face several challenges: (1) they fail to fully harness direct inter-modal cues, possibly leading to less-than-thorough cross-modal modeling; (2) they concurrently extract information from the same and different modalities at each network layer, potentially triggering conflicts from the fusion of multi-source data; (3) they lack the agility required to detect dynamic sentimental changes, perhaps resulting in inaccurate classification of utterances with abrupt sentiment shifts. To address these issues, a novel approach named GraphSmile is proposed for tracking intricate emotional cues in multimodal dialogues. GraphSmile comprises two key components, i.e., GSF and SDP modules. GSF ingeniously leverages graph structures to alternately assimilate inter-modal and intra-modal emotional dependencies layer by layer, adequately capturing cross-modal cues while effectively circumventing fusion conflicts. SDP is an auxiliary task to explicitly delineate the sentiment dynamics between utterances, promoting the model's ability to distinguish sentimental discrepancies. GraphSmile is effortlessly applied to multimodal sentiment analysis in conversation (MSAC), thus enabling simultaneous execution of MERC and MSAC tasks. Empirical results on multiple benchmarks demonstrate that GraphSmile can handle complex emotional and sentimental patterns, significantly outperforming baseline models.

cs.CL↗

arXiv2024

Harnessing Neural Unit Dynamics for Effective and Scalable Class-Incremental Learning

Class-incremental learning (CIL) aims to train a model to learn new classes from non-stationary data streams without forgetting old ones. In this paper, we propose a new kind of connectionist model by tailoring neural unit dynamics that adapt the behavior of neural networks for CIL. In each training session, it introduces a supervisory mechanism to guide network expansion whose growth size is compactly commensurate with the intrinsic complexity of a newly arriving task. This constructs a near-minimal network while allowing the model to expand its capacity when cannot sufficiently hold new classes. At inference time, it automatically reactivates the required neural units to retrieve knowledge and leaves the remaining inactivated to prevent interference. We name our model AutoActivator, which is effective and scalable. To gain insights into the neural unit dynamics, we theoretically analyze the model's convergence property via a universal approximation theorem on learning sequential mappings, which is under-explored in the CIL community. Experiments show that our method achieves strong CIL performance in rehearsal-free and minimal-expansion settings with different backbones.

cs.LG↗

arXiv2024

Tunable Collective Excitations in Epitaxial Perovskite Nickelates

The formation of plasmons through the collective excitation of charge density has generated intense discussions, offering insights to fundamental sciences and potential applications. While the underlying physical principles have been well-established, the effects of many-body interactions and orbital hybridization on plasmonic dynamics remain understudied. In this work, we present the observation of conventional metallic and correlated plasmons in epitaxial La1-xSrxNiO3 (LSNO) films with varying Sr doping concentrations (x = 0, 0.125, 0.25), unveiling their intriguing evolution. Unlike samples at other doping concentrations, the x = 0.125 intermediate doping sample does not exhibit the correlated plasmons despite showing high optical conductivity. Through a comprehensive experimental investigation using spectroscopic ellipsometry and X-ray absorption spectroscopy, the O2p-Ni3d orbital hybridization for LSNO with a doping concentration of x = 0.125 is found to be significantly enhanced, alongside a considerable weakening of its effective correlation U*. These factors account for the absence of correlated plasmons and the high optical conductivity observed in LSNO (0.125). Our results underscore the profound impact of orbital hybridization on the electronic structure and the formation of plasmon in strongly-correlated systems. This in turn suggest that LSNO could serve as a promising alternative material in optoelectronic devices.

cond-mat.mtrl-sci↗