arXiv subjects

Kang Li

Publications and source records attributed to Kang Li.

At least 55 records · Page 3Linked to original sources

arXiv2026

RareAlert: Aligning heterogeneous large language model reasoning for early rare disease risk screening

Missed and delayed diagnosis remains a major challenge in rare disease care. At the initial clinical encounters, physicians assess rare disease risk using only limited information under high uncertainty. When high-risk patients are not recognised at this stage, targeted diagnostic testing is often not initiated, resulting in missed diagnosis. Existing primary care triage processes are structurally insufficient to reliably identify patients with rare diseases at initial clinical presentation and universal screening is needed to reduce diagnostic delay. Here we present RareAlert, an early screening system which predict patient-level rare disease risk from routinely available primary-visit information. RareAlert integrates reasoning generated by ten LLMs, calibrates and weights these signals using machine learning, and distils the aligned reasoning into a single locally deployable model. To develop and evaluate RareAlert, we curated RareBench, a real-world dataset of 158,666 cases covering 33 Orphanet disease categories and more than 7,000 rare conditions, including both rare and non-rare presentations. The results showed that rare disease identification can be reconceptualised as a universal uncertainty resolution process applied to the general patient population. On an independent test set, RareAlert, a Qwen3-4B based model trained with calibrated reasoning signals, achieved an AUC of 0.917, outperforming the best machine learning ensemble and all evaluated LLMs, including GPT-5, DeepSeek-R1, Claude-3.7-Sonnet, o3-mini, Gemini-2.5-Pro, and Qwen3-235B. These findings demonstrate the diversity in LLM medical reasoning and the effectiveness of aligning such reasoning in highly uncertain clinical tasks. By incorporating calibrated reasoning into a single model, RareAlert enables accurate, privacy-preserving, and scalable rare disease risk screening suitable for large-scale local deployment.

cs.AI↗

arXiv2026

Orchestrating Specialized Agents for Trustworthy Enterprise RAG

Retrieval-Augmented Generation (RAG) shows promise for enterprise knowledge work, yet it often underperforms in high-stakes decision settings that require deep synthesis, strict traceability, and recovery from underspecified prompts. One-pass retrieval-and-write pipelines frequently yield shallow summaries, inconsistent grounding, and weak mechanisms for completeness verification. We introduce ADORE (Adaptive Deep Orchestration for Research in Enterprise), an agentic framework that replaces linear retrieval with iterative, user-steered investigation coordinated by a central orchestrator and a set of specialized agents. ADORE's key insight is that a structured Memory Bank (a curated evidence store with explicit claim-evidence linkage and section-level admissible evidence) enables traceable report generation and systematic checks for evidence completeness. Our contributions are threefold: (1) Memory-locked synthesis - report generation is constrained to a structured Memory Bank (Claim-Evidence Graph) with section-level admissible evidence, enabling traceable claims and grounded citations; (2) Evidence-coverage-guided execution - a retrieval-reflection loop audits section-level evidence coverage to trigger targeted follow-up retrieval and terminates via an evidence-driven stopping criterion; (3) Section-packed long-context grounding - section-level packing, pruning, and citation-preserving compression make long-form synthesis feasible under context limits. Across our evaluation suite, ADORE ranks first on DeepResearch Bench (52.65) and achieves the highest head-to-head preference win rate on DeepConsult (77.2%) against commercial systems.

cs.IR↗

arXiv2026

Evaluating the Diagnostic Classification Ability of Multimodal Large Language Models: Insights from the Osteoarthritis Initiative

Multimodal large language models (MLLMs) show promising performance on medical visual question answering (VQA) and report generation, but these generation and explanation abilities do not reliably transfer to disease-specific classification. We evaluated MLLM architectures on knee osteoarthritis (OA) radiograph classification, which remains underrepresented in existing medical MLLM benchmarks, even though knee OA affects an estimated 300 to 400 million people worldwide. Through systematic ablation studies manipulating the vision encoder, the connector, and the large language model (LLM) across diverse training strategies, we measured each component's contribution to diagnostic accuracy. In our classification task, a trained vision encoder alone could outperform full MLLM pipelines in classification accuracy and fine-tuning the LLM provided no meaningful improvement over prompt-based guidance. And LoRA fine-tuning on a small, class-balanced dataset (500 images) gave better results than training on a much larger but class-imbalanced set (5,778 images), indicating that data balance and quality can matter more than raw scale for this task. These findings suggest that for domain-specific medical classification, LLMs are more effective as interpreters and report generators rather than as primary classifiers. Therefore, the MLLM architecture appears less suitable for medical image diagnostic classification tasks that demand high certainty. We recommend prioritizing vision encoder optimization and careful dataset curation when developing clinically applicable systems.

cs.CV↗

arXiv2025

KOM: A Multi-Agent Artificial Intelligence System for Precision Management of Knee Osteoarthritis (KOA)

Knee osteoarthritis (KOA) affects more than 600 million individuals globally and is associated with significant pain, functional impairment, and disability. While personalized multidisciplinary interventions have the potential to slow disease progression and enhance quality of life, they typically require substantial medical resources and expertise, making them difficult to implement in resource-limited settings. To address this challenge, we developed KOM, a multi-agent system designed to automate KOA evaluation, risk prediction, and treatment prescription. This system assists clinicians in performing essential tasks across the KOA care pathway and supports the generation of tailored management plans based on individual patient profiles, disease status, risk factors, and contraindications. In benchmark experiments, KOM demonstrated superior performance compared to several general-purpose large language models in imaging analysis and prescription generation. A randomized three-arm simulation study further revealed that collaboration between KOM and clinicians reduced total diagnostic and planning time by 38.5% and resulted in improved treatment quality compared to each approach used independently. These findings indicate that KOM could help facilitate automated KOA management and, when integrated into clinical workflows, has the potential to enhance care efficiency. The modular architecture of KOM may also offer valuable insights for developing AI-assisted management systems for other chronic conditions.

cs.AI↗

arXiv2025

The Alignment Paradox of Medical Large Language Models in Infertility Care: Decoupling Algorithmic Improvement from Clinical Decision-making Quality

Large language models (LLMs) are increasingly adopted in clinical decision support, yet aligning them with the multifaceted reasoning pathways of real-world medicine remains a major challenge. Using more than 8,000 infertility treatment records, we systematically evaluate four alignment strategies: Supervised Fine-Tuning (SFT), Direct Preference Optimization (DPO), Group Relative Policy Optimization (GRPO), and In-Context Learning (ICL) through a dual-layer framework combining automatic benchmarks with blinded doctor-in-the-loop assessments. GRPO achieves the highest algorithmic accuracy across multiple decision layers, confirming the value of reinforcement-based optimization for structured prediction tasks. However, clinicians consistently prefer the SFT model, citing clearer reasoning processes (p = 0.035) and higher therapeutic feasibility (p = 0.019). In blinded pairwise comparisons, SFT attains the highest winning rate (51.2%), outperforming both GRPO (26.2%) and even physicians' original decisions (22.7%). These results reveal an alignment paradox: algorithmic improvements do not necessarily translate into higher clinical trust, and may diverge from human-centered preferences. Our findings highlight the need for alignment strategies that prioritize clinically interpretable and practically feasible reasoning, rather than solely optimizing decision-level accuracy.

cs.LG↗

arXiv2025

DICE: Discrete Inversion Enabling Controllable Editing for Multinomial Diffusion and Masked Generative Models

Discrete diffusion models have achieved success in tasks like image generation and masked language modeling but face limitations in controlled content editing. We introduce DICE (Discrete Inversion for Controllable Editing), the first approach to enable precise inversion for discrete diffusion models, including multinomial diffusion and masked generative models. By recording noise sequences and masking patterns during the reverse diffusion process, DICE enables accurate reconstruction and flexible editing of discrete data without the need for predefined masks or attention manipulation. We demonstrate the effectiveness of DICE across both image and text domains, evaluating it on models such as VQ-Diffusion, Paella, and RoBERTa. Our results show that DICE preserves high data fidelity while enhancing editing capabilities, offering new opportunities for fine-grained content manipulation in discrete spaces.

cs.CV↗

arXiv2025

JaxMARL-HFT: GPU-Accelerated Large-Scale Multi-Agent Reinforcement Learning for High-Frequency Trading

Agent-based modelling (ABM) approaches for high-frequency financial markets are difficult to calibrate and validate, partly due to the large parameter space created by defining fixed agent policies. Multi-agent reinforcement learning (MARL) enables more realistic agent behaviour and reduces the number of free parameters, but the heavy computational cost has so far limited research efforts. To address this, we introduce JaxMARL-HFT (JAX-based Multi-Agent Reinforcement Learning for High-Frequency Trading), the first GPU-accelerated open-source multi-agent reinforcement learning environment for high-frequency trading (HFT) on market-by-order (MBO) data. Extending the JaxMARL framework and building on the JAX-LOB implementation, JaxMARL-HFT is designed to handle a heterogeneous set of agents, enabling diverse observation/action spaces and reward functions. It is designed flexibly, so it can also be used for single-agent RL, or extended to act as an ABM with fixed-policy agents. Leveraging JAX enables up to a 240x reduction in end-to-end training time, compared with state-of-the-art reference implementations on the same hardware. This significant speed-up makes it feasible to exploit the large, granular datasets available in high-frequency trading, and to perform the extensive hyperparameter sweeps required for robust and efficient MARL research in trading. We demonstrate the use of JaxMARL-HFT with independent Proximal Policy Optimization (IPPO) for a two-player environment, with an order execution and a market making agent, using one year of LOB data (400 million orders), and show that these agents learn to outperform standard benchmarks. The code for the JaxMARL-HFT framework is available on GitHub.

q-fin.TR↗

arXiv2025

Assessing the Human-Likeness of LLM-Driven Digital Twins in Simulating Health Care System Trust

Serving as an emerging and powerful tool, Large Language Model (LLM)-driven Human Digital Twins are showing great potential in healthcare system research. However, its actual simulation ability for complex human psychological traits, such as distrust in the healthcare system, remains unclear. This research gap particularly impacts health professionals' trust and usage of LLM-based Artificial Intelligence (AI) systems in assisting their routine work. In this study, based on the Twin-2K-500 dataset, we systematically evaluated the simulation results of the LLM-driven human digital twin using the Health Care System Distrust Scale (HCSDS) with an established human-subject sample, analyzing item-level distributions, summary statistics, and demographic subgroup patterns. Results showed that the simulated responses by the digital twin were significantly more centralized with lower variance and had fewer selections of extreme options (all p<0.001). While the digital twin broadly reproduces human results in major demographic patterns, such as age and gender, it exhibits relatively low sensitivity in capturing minor differences in education levels. The LLM-based digital twin simulation has the potential to simulate population trends, but it also presents challenges in making detailed, specific distinctions in subgroups of human beings. This study suggests that the current LLM-driven Digital Twins have limitations in modeling complex human attitudes, which require careful calibration and validation before applying them in inferential analyses or policy simulations in health systems engineering. Future studies are necessary to examine the emotional reasoning mechanism of LLMs before their use, particularly for studies that involve simulations sensitive to social topics, such as human-automation trust.

cs.HC↗

arXiv2025

Reliability of Large Language Model Generated Clinical Reasoning in Assisted Reproductive Technology: Blinded Comparative Evaluation Study

Creating high-quality clinical Chains-of-Thought (CoTs) is crucial for explainable medical Artificial Intelligence (AI) while constrained by data scarcity. Although Large Language Models (LLMs) can synthesize medical data, their clinical reliability remains unverified. This study evaluates the reliability of LLM-generated CoTs and investigates prompting strategies to enhance their quality. In a blinded comparative study, senior clinicians in Assisted Reproductive Technology (ART) evaluated CoTs generated via three distinct strategies: Zero-shot, Random Few-shot (using shallow examples), and Selective Few-shot (using diverse, high-quality examples). These expert ratings were compared against evaluations from a state-of-the-art AI model (GPT-4o). The Selective Few-shot strategy significantly outperformed other strategies across all human evaluation metrics (p < .001). Critically, the Random Few-shot strategy offered no significant improvement over the Zero-shot baseline, demonstrating that low-quality examples are as ineffective as no examples. The success of the Selective strategy is attributed to two principles: "Gold-Standard Depth" (reasoning quality) and "Representative Diversity" (generalization). Notably, the AI evaluator failed to discern these critical performance differences. The clinical reliability of synthetic CoTs is dictated by strategic prompt curation, not the mere presence of examples. We propose a "Dual Principles" framework as a foundational methodology to generate trustworthy data at scale. This work offers a validated solution to the data bottleneck and confirms the indispensable role of human expertise in evaluating high-stakes clinical AI.

cs.AI↗

arXiv2025

SaFeR-VLM: Toward Safety-aware Fine-grained Reasoning in Multimodal Models

Multimodal Large Reasoning Models (MLRMs) demonstrate impressive cross-modal reasoning but often amplify safety risks under adversarial or unsafe prompts, a phenomenon we call the \textit{Reasoning Tax}. Existing defenses mainly act at the output level and do not constrain the reasoning process, leaving models exposed to implicit risks. In this paper, we propose SaFeR-VLM, a safety-aligned reinforcement learning framework that embeds safety directly into multimodal reasoning. The framework integrates four components: (I) QI-Safe-10K, a curated dataset emphasizing safety-critical and reasoning-sensitive cases; (II) safety-aware rollout, where unsafe generations undergo reflection and correction instead of being discarded; (III) structured reward modeling with multi-dimensional weighted criteria and explicit penalties for hallucinations and contradictions; and (IV) GRPO optimization, which reinforces both safe and corrected trajectories. This unified design shifts safety from a passive safeguard to an active driver of reasoning, enabling scalable and generalizable safety-aware reasoning. SaFeR-VLM further demonstrates robustness against both explicit and implicit risks, supporting dynamic and interpretable safety decisions beyond surface-level filtering. SaFeR-VLM-3B achieves average performance $70.13$ and $78.97$ on safety and helpfulness across six benchmarks, surpassing both same-scale and $>10\times$ larger models such as Skywork-R1V3-38B, Qwen2.5VL-72B, and GLM4.5V-106B. Remarkably, SaFeR-VLM-7B benefits from its increased scale to surpass GPT-5-mini and Gemini-2.5-Flash by \num{6.47} and \num{16.76} points respectively on safety metrics, achieving this improvement without any degradation in helpfulness performance. Our codes are available at https://github.com/HarveyYi/SaFeR-VLM.

cs.LG↗

arXiv2025

HyperVLA: Efficient Inference in Vision-Language-Action Models via Hypernetworks

Built upon language and vision foundation models with strong generalization ability and trained on large-scale robotic data, Vision-Language-Action (VLA) models have recently emerged as a promising approach to learning generalist robotic policies. However, a key drawback of existing VLAs is their extremely high inference costs. In this paper, we propose HyperVLA to address this problem. Unlike existing monolithic VLAs that activate the whole model during both training and inference, HyperVLA uses a novel hypernetwork (HN)-based architecture that activates only a small task-specific policy during inference, while still retaining the high model capacity needed to accommodate diverse multi-task behaviors during training. Successfully training an HN-based VLA is nontrivial so HyperVLA contains several key algorithm design features that improve its performance, including properly utilizing the prior knowledge from existing vision foundation models, HN normalization, and an action generation strategy. Compared to monolithic VLAs, HyperVLA achieves a similar or even higher success rate for both zero-shot generalization and few-shot adaptation, while significantly reducing inference costs. Compared to OpenVLA, a state-of-the-art VLA model, HyperVLA reduces the number of activated parameters at test time by $90\times$, and accelerates inference speed by $120\times$. Code is publicly available at https://github.com/MasterXiong/HyperVLA

cs.RO↗

arXiv2025

A Scenario-Driven Cognitive Approach to Next-Generation AI Memory

As artificial intelligence advances toward artificial general intelligence (AGI), the need for robust and human-like memory systems has become increasingly evident. Current memory architectures often suffer from limited adaptability, insufficient multimodal integration, and an inability to support continuous learning. To address these limitations, we propose a scenario-driven methodology that extracts essential functional requirements from representative cognitive scenarios, leading to a unified set of design principles for next-generation AI memory systems. Based on this approach, we introduce the \textbf{COgnitive Layered Memory Architecture (COLMA)}, a novel framework that integrates cognitive scenarios, memory processes, and storage mechanisms into a cohesive design. COLMA provides a structured foundation for developing AI systems capable of lifelong learning and human-like reasoning, thereby contributing to the pragmatic development of AGI.

cs.AI↗

arXiv2025

DiffOSeg: Omni Medical Image Segmentation via Multi-Expert Collaboration Diffusion Model

Annotation variability remains a substantial challenge in medical image segmentation, stemming from ambiguous imaging boundaries and diverse clinical expertise. Traditional deep learning methods producing single deterministic segmentation predictions often fail to capture these annotator biases. Although recent studies have explored multi-rater segmentation, existing methods typically focus on a single perspective -- either generating a probabilistic ``gold standard'' consensus or preserving expert-specific preferences -- thus struggling to provide a more omni view. In this study, we propose DiffOSeg, a two-stage diffusion-based framework, which aims to simultaneously achieve both consensus-driven (combining all experts' opinions) and preference-driven (reflecting experts' individual assessments) segmentation. Stage I establishes population consensus through a probabilistic consensus strategy, while Stage II captures expert-specific preference via adaptive prompts. Demonstrated on two public datasets (LIDC-IDRI and NPC-170), our model outperforms existing state-of-the-art methods across all evaluated metrics. Source code is available at https://github.com/string-ellipses/DiffOSeg .

cs.CV↗

arXiv2025

Painting the market: generative diffusion models for financial limit order book simulation and forecasting

Simulating limit order books (LOBs) has important applications across forecasting and backtesting for financial market data. However, deep generative models struggle in this context due to the high noise and complexity of the data. Previous work uses autoregressive models, although these experience error accumulation over longer-time sequences. We introduce a novel approach, converting LOB data into a structured image format, and applying diffusion models with inpainting to generate future LOB states. This method leverages spatio-temporal inductive biases in the order book and enables parallel generation of long sequences overcoming issues with error accumulation. We also publicly contribute to LOB-Bench, the industry benchmark for LOB generative models, to allow fair comparison between models using Level-2 and Level-3 order book data (with or without message level data respectively). We show that our model achieves state-of-the-art performance on LOB-Bench, despite using lower fidelity data as input. We also show that our method prioritises coherent global structures over local, high-fidelity details, providing significant improvements over existing methods on certain metrics. Overall, our method lays a strong foundation for future research into generative diffusion approaches to LOB modelling.

q-fin.TR↗

arXiv2025

Deep Research Agents: A Systematic Examination And Roadmap

The rapid progress of Large Language Models (LLMs) has given rise to a new category of autonomous AI systems, referred to as Deep Research (DR) agents. These agents are designed to tackle complex, multi-turn informational research tasks by leveraging a combination of dynamic reasoning, adaptive long-horizon planning, multi-hop information retrieval, iterative tool use, and the generation of structured analytical reports. In this paper, we conduct a detailed analysis of the foundational technologies and architectural components that constitute Deep Research agents. We begin by reviewing information acquisition strategies, contrasting API-based retrieval methods with browser-based exploration. We then examine modular tool-use frameworks, including code execution, multimodal input processing, and the integration of Model Context Protocols (MCPs) to support extensibility and ecosystem development. To systematize existing approaches, we propose a taxonomy that differentiates between static and dynamic workflows, and we classify agent architectures based on planning strategies and agent composition, including single-agent and multi-agent configurations. We also provide a critical evaluation of current benchmarks, highlighting key limitations such as restricted access to external knowledge, sequential execution inefficiencies, and misalignment between evaluation metrics and the practical objectives of DR agents. Finally, we outline open challenges and promising directions for future research. A curated and continuously updated repository of DR agent research is available at: {https://github.com/ai-agents-2030/awesome-deep-research-agent}.

cs.AI↗

arXiv2025

Maximal ideals of reduced group C*-algebras and Thompson's groups

Given a conditional expectation $P$ from a C*-algebra $B$ onto a C*-subalgebra $A$, we observe that induction of ideals via $P$, together with a map which we call co-induction, forms a Galois connection between the lattices of ideals of $A$ and $B$. Using properties of this Galois connection, we show that, given a discrete group $G$ and a stabilizer subgroup $G_x$ for the action of $G$ on its Furstenberg boundary, induction gives a bijection between the set of maximal co-induced ideals of $C^*(G_x)$ and the set of maximal ideals of $C^*_r(G)$. As an application, we prove that the reduced C*-algebra of Thompson's group $T$ has a unique maximal ideal. Furthermore, we show that, if Thompson's group $F$ is amenable, then $C^*_r(T)$ has infinitely many ideals.

math.OA↗

arXiv2025

Multi-Robot Cooperative Herding through Backstepping Control Barrier Functions

We propose a novel cooperative herding strategy through backstepping control barrier functions (CBFs), which coordinates multiple herders to herd a group of evaders safely towards a designated goal region. For the herding system with heterogeneous groups involving herders and evaders, the behavior of the evaders can only be influenced indirectly by the herders' motion, especially when the evaders follow an inverse dynamics model and respond solely to repulsive interactions from the herders. This indirect interaction mechanism inherently renders the overall system underactuated. To address this issue, we first construct separate CBFs for the dual objectives of goal reaching and collision avoidance, which ensure both herding completion and safety guarantees. Then, we reformulate the underactuated herding dynamics into a control-affine structure and employ a backstepping approach to recursively design control inputs for the hierarchical barrier functions, avoiding taking derivatives of the higher-order system. Finally, we present a cooperative herding strategy based on backstepping CBFs that allow herders to safely herd multiple evaders into the goal region. In addition, centralized and decentralized implementations of the proposed algorithm are developed, further enhancing its flexibility and applicability. Extensive simulations and real-world experiments validate the effectiveness and safety of the proposed strategy in multi-robot herding.

cs.MA↗

arXiv2025

Attention-Guided Multi-Scale Local Reconstruction for Point Clouds via Masked Autoencoder Self-Supervised Learning

Self-supervised learning has emerged as a prominent research direction in point cloud processing. While existing models predominantly concentrate on reconstruction tasks at higher encoder layers, they often neglect the effective utilization of low-level local features, which are typically employed solely for activation computations rather than directly contributing to reconstruction tasks. To overcome this limitation, we introduce PointAMaLR, a novel self-supervised learning framework that enhances feature representation and processing accuracy through attention-guided multi-scale local reconstruction. PointAMaLR implements hierarchical reconstruction across multiple local regions, with lower layers focusing on fine-scale feature restoration while upper layers address coarse-scale feature reconstruction, thereby enabling complex inter-patch interactions. Furthermore, to augment feature representation capabilities, we incorporate a Local Attention (LA) module in the embedding layer to enhance semantic feature understanding. Comprehensive experiments on benchmark datasets ModelNet and ShapeNet demonstrate PointAMaLR's superior accuracy and quality in both classification and reconstruction tasks. Moreover, when evaluated on the real-world dataset ScanObjectNN and the 3D large scene segmentation dataset S3DIS, our model achieves highly competitive performance metrics. These results not only validate PointAMaLR's effectiveness in multi-scale semantic understanding but also underscore its practical applicability in real-world scenarios.

cs.GR↗