Recent advances in multimodal large language models (MLLMs) have demonstrated remarkable multi- and cross-modal integration capabilities. However, their potential for fine-grained emotion understanding remains systematically underexplored. While open-vocabulary multimodal emotion recognition (MER-OV) has emerged as a promising direction to overcome the limitations of closed emotion sets, no comprehensive evaluation of MLLMs in this context currently exists. To address this, our work presents the first large-scale benchmarking study of MER-OV on the OV-MERD dataset, evaluating 19 mainstream MLLMs, including general-purpose, modality-specialized, and reasoning-enhanced architectures. Through systematic analysis of model reasoning capacity, fusion strategies, contextual utilization, and prompt design, we provide key insights into the capabilities and limitations of current MLLMs for MER-OV. Our evaluation reveals that a two-stage, trimodal (audio, video, and text) fusion approach achieves optimal performance in MER-OV, with video emerging as the most critical modality. We further identify a surprisingly narrow gap between open- and closed-source LLMs. These findings establish essential benchmarks and offer practical guidelines for advancing open-vocabulary and fine-grained affective computing, paving the way for more nuanced and interpretable emotion AI systems. Associated code will be made publicly available upon acceptance.
The emergence of large language models (LLMs) has significantly advanced natural language understanding, generation, and reasoning across various domains, including the biomedical field. Despite these advancements, the evaluation of biomedical LLMs remains limited, primarily relying on manually crafted datasets which are insufficient for comprehensively assessing LLMs’ capabilities. To address this challenge and cater specifically to the requirements of biomedical LLMs, we propose KGMedQA, an innovative evaluation benchmark based on knowledge graphs (KGs) designed to assess the knowledge and reasoning abilities of LLMs. Through careful alignment between natural language and KG structures, KGMedQA can be applied to arbitrary KGs, enabling automated question/answer generation and LLM evaluation. By leveraging the advantages of KG structures, we design seven tasks of varying complexity and focus, accompanied by specialized evaluation metrics. Experiments conducted with KGMedQA involve ten different LLMs, including general and specialized biomedical models, tested across two KGs focusing on different types of biomedical knowledge and reasoning. Compared to traditional methods, our results uncover more novel insights. For instance, while specialized models exhibit strengths in knowledge, they have deficiencies in reasoning abilities compared to general models. Additionally, factors such as model scales and prompting methods also impact the performance of LLMs. Our benchmark represents advancements in the evaluation of domain-specific LLMs, offering an effective tool for future research and development. Our source code is available at: https://github.com/PerseidsMeteorShower/KGMedQA.
While rapid advances in large language models (LLMs) are reshaping data-driven intelligent education, accurately simulating students remains an important but challenging bottleneck for scalable educational data collection, evaluation, and intervention design. However, current works are limited by scarce real interaction data, costly expert evaluation for realism, and a lack of large-scale, systematic analyses of LLMs ability in simulating students. We address this gap by presenting a three-stage LLM-human collaborative pipeline to automatically generate and filter high-quality student agents. We leverage a two-round automated scoring validated by human experts and deploy a score propagation module to obtain more consistent scores across the student similarity graph. Experiments show that combining automated scoring, expert calibration, and graph-based propagation yields simulated student that more closely track authentication by human judgments. We then analyze which profiles and behaviors are simulated more faithfully, supporting subsequent studies on personalized learning and educational assessment.
Project compliance review serves as a critical component in ensuring that project submissions meet regulatory and procedural standards while filtering out non-compliant proposals. However, the rapid increase in project applications has rendered traditional manual review mechanisms inefficient and unsustainable. Existing automated approaches—based on either semantic vector matching or direct use of large language models (LLMs)—often struggle with heterogeneous document structures, limited robustness, and high computational cost. To overcome these challenges, this study proposes an automated compliance review framework based on multi-agent routing. The framework integrates multiple expert agents with distinct reasoning paradigms and employs a classification agent to dynamically route each task to the most suitable expert, thereby enabling data-adaptive decision-making. Through adaptive routing, the system effectively reduces redundant computation by invoking complex reasoning only when necessary. Experimental results on real-world datasets from the power industry demonstrate that our method achieves superior accuracy and 48% reduction in token consumption compared to the best-performing single-agent baseline, achieving a balanced trade-off between effectiveness and efficiency. Furthermore, an online review platform has been developed and successfully deployed in large-scale power project evaluation scenarios, validating the practicality and scalability of the proposed approach.
Precise vehicle control at the limits of tire adhesion is paramount for both competitive motorsport performance and the safe execution of emergency maneuvers in road vehicles. Mastering this “grip-limit driving” presents significant challenges due to highly non-linear vehicle dynamics and sensitivity to changing conditions, often exceeding the capabilities of traditional controllers and driver models. This paper investigates the efficacy of Deep Reinforcement Learning (DRL), specifically the Proximal Policy Optimisation (PPO) algorithm, as a data-driven approach to learn expert-level driving skills within the TORCS high-fidelity race car simulation environment. An agent was trained end-to-end, utilizing “realworld-friendly” state signals (such as speeds, accelerations, and yaw rate, simple LiDaR, etc.) as input to determine continuous steering and pedal commands. Notably, the trained Agent achieved lap times comparable to a human e-sport world champion on the target track, demonstrating the potential of this methodology while also highlighting how agents can exploit idealized simulation to achieve superhuman control. Furthermore, this work presents the formulation of the time-optimal driving task as a DRL problem and offers a novel justification for the commonly used “progress reward” function, demonstrating its conceptual link to the time-difference feedback mechanisms human drivers use for performance optimization. These findings provide valuable insights into AI-driven vehicle control under extreme conditions and contribute to the development of more capable autonomous agents for simulation and potentially, real-world applications.
Electronic Health Records (EHRs) continuously monitor patients’ health status in Intensive Care Units (ICUs), capturing irregular numerical time-series data and unstructured clinical text. While existing studies primarily focus on handling modality irregularities, they often overlook the complex intra- and inter-sequence interactions as well as the dependencies between short-term and long-term features. Moreover, clinical notes are typically semantically sparse and structurally noisy, making them difficult to interpret. To address these challenges, we propose a novel multimodal predictive model. For irregular numerical time-series data, we design a cross-view multi-scale framework that integrates cross-attention mechanisms with multi-scale convolutions. This enables dynamic modeling of diverse temporal embeddings while precisely capturing intrinsic inter-variable interactions and cross-temporal dependencies, all with reduced computational complexity. For clinical text, we adopt a retrieval-augmented technique that leverages external medical knowledge graphs (KGs) and large language models (LLMs) to enrich text representations related to medical codes. These enhanced embeddings are then fused with clinical notes via a gated mechanism, effectively alleviating semantic sparsity. We validate the effectiveness of the proposed approach on two critical clinical prediction tasks. Experimental results show maximum relative F1 score improvements of 3.3%, 6.0%, and 3.4% for MISTS, clinical notes, and multimodal fusion tasks, respectively, demonstrating our method’s excellent medical predictive capability.
Large Language Models (LLMs) have exhibited impressive capabilities in reasoning and language understanding. However, their reliance on memorized knowledge and tendency to generate hallucinated content limit their reliability in real-world applications. Retrieval-Augmented Generation (RAG) mitigates these issues by integrating a retrieval module that supplements LLMs with relevant external knowledge. This paradigm bridges parametric memory and explicit retrieval, offering a principled way to ground generation in factual evidence. Despite substantial progress, most prior work has focused on optimizing isolated components, either retrieval or generation, while overlooking the agentic perspective, in which LLMs act as autonomous agents capable of actively acquiring and strategically utilizing knowledge. In this perspectives paper, we argue for reinterpreting RAG as a collaborative knowledge process among agents with distinct yet complementary roles. We categorize knowledge-intensive agents into two primary roles: knowledge acquisition (e.g., routing, query reformulation) and knowledge utilization (e.g., knowledge refinement, response generation). From this viewpoint, RAG becomes a dynamic system in which knowledge is continuously transmitted, transformed, and aligned across agent roles. To fully realize this paradigm, we advocate a joint optimization framework for knowledge-intensive agents within RAG systems. This framework explicitly models the dynamics of knowledge flow in multi-agent settings, aligning knowledge supply with knowledge demand through LLM-driven data synthesis, feedback, and evaluation. By fostering adaptive and targeted knowledge exchange, the framework mitigates conflicts between parametric and retrieved knowledge, thereby enhancing both coherence and factuality. We argue that this multi-agent joint optimization paradigm improves RAG systems in scalability, reliability, and adaptability, unlocking the potential for next-generation knowledge-intensive LLMs that reason, retrieve, and collaborate across deep retrieval processes and diverse vertical domains.
While world models have emerged as a cornerstone of embodied intelligence by enabling agents to reason about environmental dynamics through action-conditioned prediction, their evaluation remains fragmented. Current evaluation of embodied world models has largely focused on perceptual fidelity (e.g., video generation quality), overlooking the functional utility of these models in downstream decision-making tasks. In this work, we introduce WorldArena, a unified benchmark designed to systematically evaluate embodied world models across both perceptual and functional dimensions. WorldArena assesses models through three dimensions: video perception quality, measured with 16 metrics across six sub-dimensions; embodied task functionality, which evaluates world models as data engines, policy evaluators, and action planners integrating with subjective human evaluation. Furthermore, we propose EWMScore, a holistic metric integrating multi-dimensional performance into a single interpretable index. Through extensive experiments on 14 representative models, we reveal a significant perception–functionality gap, showing that high visual quality does not necessarily translate into strong embodied task capability. WorldArena benchmark with the public leaderboard is released at https://world-arena.ai/, providing a framework for tracking progress toward truly functional world models in embodied AI.
Recommender systems serve as a foundational infrastructure in modern information ecosystems, helping users navigate the expanding digital content space and discover items aligned with their preferences. At their core, recommender systems address a fundamental research problem: matching users with items. Over the past decades, the field has experienced successive technological paradigm shifts, from collaborative filtering and matrix factorization in the machine learning era to sophisticated neural architectures in the deep learning era. Recently, the emergence of generative models, especially large language models (LLMs) and diffusion models have sparked a new paradigm: generative recommendation, which reconceptualizes the recommendation problem as a generation task rather than a discriminative scoring procedure. This survey provides a comprehensive examination of this paradigm through a unified tripartite framework spanning data, model, and task dimensions. Rather than simply categorizing works, we systematically decompose approaches into operational stages—data augmentation and unification, model alignment and training, task formulation and execution. At the data level, generative models enable knowledge-infused augmentation and agent-based simulation while unifying heterogeneous signals. At the model level, we taxonomize LLM-based methods, large recommendation models, and diffusion approaches, analyzing their alignment mechanisms and innovations. At the task level, we illuminate new capabilities including conversational interaction, explainable reasoning, and personalized content generation. We identify five key advantages: world knowledge integration, natural language understanding, reasoning capabilities, scaling laws, and creative generation. We critically examine challenges in benchmark design, model robustness, and deployment efficiency, while charting a roadmap toward intelligent recommendation assistants that fundamentally reshape human-information interaction.
Agentic AI systems, built upon large language models (LLMs) and deployed in multi-agent configurations, are redefining intelligence, autonomy, collaboration, and decision-making across enterprise and societal domains. This review presents a structured analysis of Trust, Risk, and Security Management (TRiSM) in the context of LLM-based Agentic Multi-Agent Systems (AMAS). We begin by examining the conceptual foundations of Agentic AI and highlight its architectural distinctions from traditional AI agents. We then adapt and extend the AI TRiSM framework for Agentic AI, structured around key pillars: Explainability, ModelOps, Security, Privacy and their Lifecycle Governance, each contextualized to the challenges of AMAS. A risk taxonomy is proposed to capture the unique threats and vulnerabilities of Agentic AI, ranging from coordination failures to prompt-based adversarial manipulation. To make coordination and tool use measurable in practice, we propose two metrics: the Component Synergy Score (CSS), which captures inter-agent enablement, and the Tool Utilization Efficacy (TUE), which evaluates whether tools are invoked correctly and efficiently. We further discuss strategies for improving explainability in Agentic AI, as well as approaches to enhancing security and privacy through encryption, adversarial robustness, and regulatory compliance. The review concludes with a research roadmap for the responsible development and deployment of Agentic AI, highlighting key directions to align emerging systems with TRiSM principles-ensuring safety, transparency, and accountability in their operation.
While the task of face swapping has recently gained attention in the research community, a related problem of head swapping remains largely unexplored. In addition to skin color transfer, head swap poses extra challenges, such as the need to preserve structural information of the whole head during synthesis and inpaint gaps between swapped head and background. In this paper, we address these concerns with GHOST 2.0, which consists of two problem-specific modules. First, we introduce enhanced Aligner model for head reenactment, which preserves identity information at multiple scales and is robust to extreme pose variations. Secondly, we use a Blender module that seamlessly integrates the reenacted head into the target background by transferring skin color and inpainting mismatched regions. Both modules outperform the baselines on the corresponding tasks, allowing to achieve state of the art results in head swapping. We also tackle complex cases, such as large difference in hair styles of source and target.
Traditional neural networks employ fixed weights during inference, limiting their ability to adapt to changing input conditions, unlike biological neurons that adjust signal strength dynamically based on stimuli. This discrepancy between artificial and biological neurons constrains neural network flexibility and adaptability. To bridge this gap, we propose a novel framework for adaptive neural networks, where neuron weights are modeled as functions of the input signal, allowing the network to adjust dynamically in real-time. Importantly, we achieve this within the same traditional architecture of an Artificial Neural Network, maintaining structural familiarity while introducing dynamic adaptability. In our research, we apply Chebyshev polynomials as one of the many possible decomposition methods to achieve this adaptive weighting mechanism, with polynomial coefficients learned during training. Of the 145 datasets tested, our adaptive Chebyshev neural network demonstrated a marked improvement over an equivalent MLP in approximately 83% of the cases, performing strictly better on 121 datasets. In the remaining 24 datasets, the performance of our algorithm matched that of the MLP, highlighting its ability to generalize the behavior of standard neural networks while offering enhanced adaptability. As a generalized form of MLP, this model seamlessly retains MLP performance where needed while extending its capabilities to achieve superior accuracy across a wide range of complex tasks. These results underscore the potential of adaptive neurons to enhance generalization, flexibility, and robustness in neural networks, particularly in applications with dynamic or non-linear data dependencies.
Handling missing values in tabular datasets presents a significant challenge in training and testing artificial intelligence models, an issue usually addressed using imputation techniques. Here we introduce "Not Another Imputation Method" (NAIM), a novel transformer-based model specifically designed to address this issue without the need for traditional imputation techniques. NAIM's ability to avoid the necessity of imputing missing values and to effectively learn from available data relies on two main techniques: the use of feature-specific embeddings to encode both categorical and numerical features also handling missing inputs; the modification of the masked self-attention mechanism to completely mask out the contributions of missing data. Additionally, a novel regularization technique is introduced to enhance the model's generalization capability from incomplete data. We extensively evaluated NAIM on 5 publicly available tabular datasets, demonstrating its superior performance over 6 state-of-the-art machine learning models and 5 deep learning models, each paired with 3 different imputation techniques when necessary. The results highlight the efficacy of NAIM in improving predictive performance and resilience in the presence of missing data. To facilitate further research and practical application in handling missing data without traditional imputation methods, we made the code for NAIM available at https://github.com/cosbidev/NAIM.
Scientific research is not a linear pipeline but a dynamic system built upon the ever-shifting interactions among three elements — research objects, tools, and researchers. And sustained progress depends on how quickly insights circulate within this network, not on optimizing a single node in isolation. With the impending arrival of more general artificial intelligence, we stand at a critical point in how AI might change scientific research in a systemic manner. Recent “AI for Science” achievements—from protein-structure prediction to accelerated climate simulations—have proven the value of task-level AI-driven solutions. Yet, potential still remains unrealized when these advances are siloed in disciplinary “archipelagos”. This paper argues that the real prize is systemic: AI that simultaneously expands the research objects’ data landscape (AI for Data), rewires computational research tools (AI for Computation), and co-creates hypotheses with researchers (AI for Innovation). When these three pushes converge, AI stops being merely a revolution of tools but becomes the tool of revolution—a catalyst that raises the frequency, breadth, and depth of discovery across disciplines. By enhancing the full research triad rather than isolated nodes, AI can raise the overall tempo and scope of discovery in a measured, discipline-agnostic way.
Recent advances in large language models have transformed urban planning from passive tool-assisted workflows to active human–AI collaborative partnerships, enabling natural language-driven design generation, multi-agent stakeholder simulation, and intelligent decision support. This survey systematically examines the integration of LLMs in urban planning, establishing a comprehensive taxonomy covering task categories, technical paradigms, and collaboration patterns. Furthermore, the survey identifies critical evaluation frameworks and benchmark datasets while examining implementation challenges, including domain knowledge integration, scalability constraints, and ethical implications. The work bridges theoretical advances with practical deployment considerations, providing guidance for selecting appropriate LLM approaches across different urban planning contexts and scales.
Knowledge graphs (KGs) encode structured information about real-world entities and their relations, supporting core NLP tasks such as question answering and retrieval. Existing LLM-based methods for knowledge extraction and fusion often struggle to balance quality and coverage when adapting to emerging knowledge. We propose LLMKG＋, a framework for KG expansion that integrates the generative strengths of LLMs with relevance verification. LLMKG＋features (1) a two-stage pipeline with retrieval-augmented generation followed by hierarchical expansion filtering, where the latter is the first to jointly assess semantic equivalence to eliminate triple-level redundancy while ensuring factual correctness, and (2) a novel KG Reconstruction Test that recognizes semantically equivalent triples to enable more accurate quality and coverage assessment. Evaluated on PubMed abstracts and the UMLS semantic network using eight state-of-the-art LLMs, LLMKG＋improves KG quality and coverage by 20.47%–73.71% over strong baselines. These results demonstrate that LLMKG＋offers an effective solution for KG expansion in domains requiring high quality, broad coverage, and continual knowledge growth. Code: https://github.com/xincanfeng/llmkg.
Emotion recognition has become increasingly significant in artificial intelligence; however, the impact of body movements on emotion interpretation remains under-explored. This paper presents a novel Hybrid Bayesian Pre-trained Long Short-Term Memory (HBP-LSTM) framework that combines low-level pose data with high-level kinematic features, utilising Bayesian inference to enhance the accuracy and robustness of emotion recognition. The proposed model is trained on high-quality laboratory data to capture the fundamental patterns of emotional expression through body movements. We introduce noise and employ adversarial attack methods such as the Fast Gradient Sign Method (FGSM) to evaluate the model’s robustness during testing. This approach assesses the HBP-LSTM’s ability to maintain performance under data degradation and adversarial conditions, common challenges in real-world scenarios. We validated the HBP-LSTM on two public datasets, EGBM and KDAEE, demonstrating that the model exhibits high robustness against noise and adversarial perturbations, outperforming traditional models. The HBP-LSTM accurately identifies seven basic emotions (happiness, sadness, surprise, fear, anger, disgust, and neutrality) with accuracies of 98% and 88% on the EGBM and KDAEE datasets, respectively. HBP-LSTM is a noise-resistant model with a reliable emotion recognition framework, which lays the foundation for future applications of emotion recognition technology in more challenging real-world environments.
Intelligent speech analysis (ISA) constitutes a significant component within the realm of computer audition (CA) technology. Speech, as a fundamental tool for human communication, not only conveys rich semantic information but also holds significant potential for various healthcare applications. Computational paralinguistics methods can be used to analyse alterations in the acoustic characteristics of speech signals induced by medical conditions, providing valuable insights into shifts in an individual’s health status. More importantly, compared to other physiological monitoring devices, speech acquisition devices are non-invasive and user-friendly, making them accessible for a wide range of individuals. However, despite its promise, ISA in healthcare currently faces a range of notable challenges that hinder its widespread adoption. In this survey, we present an overview of the development and current research in speech analysis technologies within the healthcare domain. First, we summarise the methodologies employed in ISA-based healthcare. Next, we provide an overview of applications in evaluating physical diseases, mental health conditions, and neurological disorders. Additionally, we discuss key limitations and shortcomings in the current state of the field. Finally, we conclude with a summary of the discussed works and offer insights into future research directions aimed at addressing these limitations to advance the practical implementation of ISA in clinical settings. This survey aims to serve as a valuable resource for researchers in speech analysis, biomedicine, and related fields. We hope to inspire greater interest in this promising area within the scientific community and provide guidance for future studies in this evolving field.
In recent years, GCN-based methods have achieved great success in skeleton-based human motion prediction tasks due to the human body graph structure. However, existing methods leveraged single semantic information to model the whole motion sequence, which cannot fully exploit the motion dependencies. To tackle this issue, we propose a Multi-spatial Semantic Information Aggregation Network(MSIAN) to enrich the semantic information by focusing on the local spatial structure of the human skeleton. MSIAN includes the Graph-based Feature Extraction and Aggregation Block (GFEAB), where the Integration Graph combines local and global attention to extract spatial features, the Gravity-Centered Graph (GCG) captures the state of each joint by treating the central joint of the skeleton as the center of gravity, and the Spatial Position Graph (SPG) fully utilizes the original joint positions to analyze movements. Extensive experiments show that our proposed MSIAN outperforms the current state-of-the-art methods on Human3.6M, 3DPW, and AMASS datasets. Our code is available at https://github.com/HDdong-hub/MSIAN.
Accurate motion forecasting is essential for the safety and reliability of autonomous driving (AD) systems. While existing methods have made significant progress, they often overlook explicit safety constraints and struggle to capture the complex interactions among traffic agents, environmental factors, and motion dynamics. To address these challenges, we present SafeCast, a risk-responsive motion forecasting model that integrates safety-aware decision-making with uncertainty-aware adaptability. SafeCast is the first to incorporate the Responsibility-Sensitive Safety (RSS) framework into motion forecasting, encoding interpretable safety rules — such as safe distances and collision avoidance — based on traffic norms and physical principles. To further enhance robustness, we introduce the Graph Uncertainty Feature (GUF), a graph-based module that injects learnable noise into Graph Attention Networks, capturing real-world uncertainties and enhancing generalization across diverse scenarios. We evaluate SafeCast on four real-world benchmark datasets — Next Generation Simulation (NGSIM), Highway Drone (HighD), ApolloScape, and the Macao Connected Autonomous Driving (MoCAD) — covering highway, urban, and mixed-autonomy traffic environments. Our model achieves state-of-the-art (SOTA) accuracy while maintaining a lightweight architecture and low inference latency, underscoring its potential for real-time deployment in safety-critical AD systems.
