arXiv subjects

Mengran Li

Publications and source records attributed to Mengran Li.

At least 19 recordsLinked to original sources

arXiv2026

Do Pathology Vision-Language Models Truly See Pathology?

Pathology vision-language models (VLMs) have recently progressed rapidly and are commonly evaluated by answer accuracy on pathology VQA benchmarks. However, we dig into current evaluations and identify three overlooked issues: 1) Visual evidence is not always necessary. For instance, Gemini-3-Pro achieves 53.5% average accuracy across 5 VQA benchmarks without any visual input. 2) Domain training can improve accuracy without proportional gains in visual binding. Compared with Qwen2.5-VL-7B, Patho-R1-7B exhibits a 5.8-point lower multimodal gain and a 3.7-point lower attention IoU. 3) Entity-level attention is diffuse and weakly query-specific. On PathVG, attention maps remain highly correlated across different entity queries. These issues can lead to substantial misjudgments of pathology VLMs' actual multimodal capabilities. To this end, we present PathBind, a benchmark comprising 2,600 samples: PathBind-VQA with 1,500 questions across six dimensions, PathBind-PTA with 600 questions from a private pathology teaching atlas, and PathBind-Grounding with 500 expert-curated region-level samples. Each component undergoes task-specific automated filtering and expert review to reduce textual shortcuts and improve entity-region correspondence. We evaluate 18 representative VLMs on VQA samples of PathBind and five existing pathology VQA benchmarks, and further evaluate 10 VLMs on PathBind-Grounding and PathVG. Results show that current pathology VLMs still exhibit a substantial gap between answer-side performance and visual-semantic binding.

cs.CV↗

arXiv2026

PathoSage: Towards Multi-Source Evidence Adjudication in Pathology via Experience-Aware Agentic Workflow

Recent advances in Multimodal Large Language Models (MLLMs) and agent workflows have shown strong promise for computational pathology, yet reliable patch-level reasoning remains challenging. End-to-end pathology MLLMs often hallucinate morphological features, while recent agentic systems usually merge tool outputs and retrieved knowledge into a shared context, making decisions vulnerable to conflicting evidence and context contamination. We propose PathoSage, a three-stage framework that explicitly separates knowledge retrieval, evidence collection, and evidence adjudication for patch-level pathology multimodal reasoning. Its core component, Structured Evidence Deliberation, independently evaluates heterogeneous evidence from tools, performs conflict analysis, and generates the final judgment in a fresh context to reduce anchoring bias. We further introduce a training-free Beta-Bernoulli experience system with continuous credit assignment to model long-term tool reliability and construct similarity-weighted priors for future tool use. Experiments show that PathoSage effectively mitigates VQA hallucinations and classifier disagreement, outperforming strong pathology MLLM and agentic baselines. Our results highlight explicit evidence adjudication and reliability-aware tool modeling as key ingredients for robust pathology agents.

cs.AI↗

arXiv2026

CellScientist: Dual-Space Hierarchical Orchestration for Closed-Loop Refinement of Virtual Cell Models

Virtual Cell Modeling (VCM) requires models that not only predict perturbation responses, but also support targeted revision when predictions fail. Current LLM-assisted modeling workflows face a refinement-routing problem: prediction discrepancies are observed through executable implementations, but the relevant revision may involve the modeling assumption, representation design, implementation, or task constraint. Without structured feedback propagation across these levels, iterative refinement may repair code while failing to revise the assumption responsible for the discrepancy. We propose CellScientist, a dual-space hierarchical framework that couples a high-level hypothesis space with a low-level executable implementation space. CellScientist represents modeling decisions as structured states, realizes them as admissible programs under task and interface constraints, and routes execution discrepancies back to targeted hypothesis or implementation updates. This enables a closed Hypothesis -> Implementation -> Hypothesis loop where failures become structured signals for model refinement rather than debugging events. Across morphology and transcriptomic benchmarks, with additional single-cell perturbation evaluations, the final executable models selected by CellScientist improve over reference baselines under fixed split and evaluation protocols, while the workflow produces auditable refinement traces.

cs.LG↗

arXiv2026

Causal Discovery in Multivariate Extremes via Tail Asymmetry

Causal discovery in multivariate extremes is challenging because extreme observations are sparse, dependent, and often affected by latent common shocks. Existing approaches focus on undirected extremal dependence, require prior graph restriction, and do not scale beyond small systems. We introduce tail-induced asymmetry as a principle for causal directionality in heavy-tailed systems, where extreme events propagate asymmetrically so that forward tail prediction is systematically easier than backward prediction. We show that this asymmetry yields identifiable causal direction under a canonical max-linear model and provides a basis for score-based structure learning in the tail regime. Building on this, we propose Sparse Structure diScovery in Multivariate Extremes (S3ME), a two-stage data-driven framework for causal discovery. The first stage performs proxy-adjusted penalized neighbourhood selection to recover a sparse candidate skeleton under latent confounding. The second stage orients edges by minimizing tail prediction risk based on max-linear envelope models, exploiting directional asymmetry. We establish high-dimensional guarantees for skeleton screening and consistency of the score-based estimator under population separation conditions. Simulations demonstrate robustness to latent confounding and favourable scaling relative to existing extremal methods. Applications to river network data and financial tail-risk networks show that the approach recovers sparse, interpretable propagation structures without prespecified graph structure.

stat.ME↗

arXiv2026

Tail-Calibrated Estimation of Extreme Quantile Treatment Effects

Extreme quantile treatment effects (eQTEs) measure the causal impact of a treatment on the tails of an outcome distribution and are central for studying rare, high-impact events. Standard QTE methods often fail in extreme regimes due to data sparsity, while existing eQTE methods rely on restrictive tail assumptions or on interior-quantile theory. We propose the Tail-Calibrated Inverse Estimating Equation (TIEE) framework, which combines information across quantile levels and anchors the tail using extreme value models within a unified estimating equation approach. We establish asymptotic properties of the resulting estimator and evaluate its performance through simulation under different tail behaviours and model misspecifications. An application to extreme precipitation in the Austrian Alps illustrates how TIEE enables observational causal attribution for very rare events under anthropogenic warming. More broadly, the proposed framework establishes a new foundation for causal inference on rare, high-impact outcomes, with relevance across environmental risk, economics, and public health.

stat.ME↗

arXiv2026

Learning to Decode Against Compositional Hallucination in Video Multimodal Large Language Models

Current research on video hallucination mitigation primarily focuses on isolated error types, leaving compositional hallucinations, arising from incorrect reasoning over multiple interacting spatial and temporal factors largely underexplored. We introduce OmniVCHall, a benchmark designed to systematically evaluate both isolated and compositional hallucinations in video multimodal large language models (VLLMs). OmniVCHall spans diverse video domains, introduces a novel camera-based hallucination type, and defines a fine-grained taxonomy, together with adversarial answer options (e.g., "All are correct" and "None of the above") to prevent shortcut reasoning. The evaluations of 39 representative VLLMs reveal that even advanced models (e.g., Qwen3-VL and GPT-5) exhibit substantial performance degradation. We propose TriCD, a contrastive decoding framework with a triple-pathway calibration mechanism. An adaptive perturbation controller dynamically selects distracting operations to construct negative video variants, while a saliency-guided enhancement module adaptively reinforces grounded token-wise visual evidences. These components are optimized via reinforcement learning to encourage precise decision-making under compositional hallucination settings. Experimental results show that TriCD consistently improves performance across two representative backbones, achieving an average accuracy improvement of over 10%. The data and code can be find at https://github.com/BMRETURN/OmniVCHall.

cs.CV↗

arXiv2025

Learning Cell-Aware Hierarchical Multi-Modal Representations for Robust Molecular Modeling

Understanding how chemical perturbations propagate through biological systems is essential for robust molecular property prediction. While most existing methods focus on chemical structures alone, recent advances highlight the crucial role of cellular responses such as morphology and gene expression in shaping drug effects. However, current cell-aware approaches face two key limitations: (1) modality incompleteness in external biological data, and (2) insufficient modeling of hierarchical dependencies across molecular, cellular, and genomic levels. We propose CHMR (Cell-aware Hierarchical Multi-modal Representations), a robust framework that jointly models local-global dependencies between molecules and cellular responses and captures latent biological hierarchies via a novel tree-structured vector quantization module. Evaluated on nine public benchmarks spanning 728 tasks, CHMR outperforms state-of-the-art baselines, yielding average improvements of 3.6% on classification and 17.2% on regression tasks. These results demonstrate the advantage of hierarchy-aware, multimodal learning for reliable and biologically grounded molecular representations, offering a generalizable framework for integrative biomedical modeling. The code is in https://github.com/limengran98/CHMR.

cs.LG↗

arXiv2025

On the importance of tail assumptions in climate extreme event attribution

Extreme weather events are becoming more frequent and intense, posing serious threats to human life, biodiversity, and ecosystems. A key objective of extreme event attribution (EEA) is to assess whether and to what extent anthropogenic climate change influences such events. Central to EEA is the accurate statistical characterization of atmospheric extremes, which are inherently multivariate or spatial due to their measurement over high-dimensional grids. Within the counterfactual causal inference framework of Pearl, we evaluate how tail assumptions affect attribution conclusions by comparing three multivariate modeling approaches for estimating causation metrics. These include: (i) the multivariate generalized Pareto distribution, which imposes an invariant tail dependence structure; (ii) the factor copula model of Castro-Camilo and Huser (2020), which offers flexible subasymptotic behavior; and (iii) the model of Huser and Wadsworth (2019), which smoothly transitions between different forms of extremal dependence. We assess the implications of these modeling choices in both simulated scenarios (under varying forms of model misspecification) and real data applications, using weekly winter maxima over Europe from the M\'et\'eo-France CNRM model and daily precipitation from the ACCESS-CM2 model over the U.S. Our findings highlight that tail assumptions critically shape causality metrics in EEA. Misspecification of the extremal dependence structure can lead to substantially different and potentially misleading attribution conclusions, underscoring the need for careful model selection and evaluation when quantifying the influence of climate change on extreme events.

stat.AP↗

arXiv2025

A Survey of Large Language Models for Data Challenges in Graphs

Graphs are a widely used paradigm for representing non-Euclidean data, with applications ranging from social network analysis to biomolecular prediction. While graph learning has achieved remarkable progress, real-world graph data presents a number of challenges that significantly hinder the learning process. In this survey, we focus on four fundamental data-centric challenges: (1) Incompleteness, real-world graphs have missing nodes, edges, or attributes; (2) Imbalance, the distribution of the labels of nodes or edges and their structures for real-world graphs are highly skewed; (3) Cross-domain Heterogeneity, graphs from different domains exhibit incompatible feature spaces or structural patterns; and (4) Dynamic Instability, graphs evolve over time in unpredictable ways. Recently, Large Language Models (LLMs) offer the potential to tackle these challenges by leveraging rich semantic reasoning and external knowledge. This survey focuses on how LLMs can address four fundamental data-centric challenges in graph-structured data, thereby improving the effectiveness of graph learning. For each challenge, we review both traditional solutions and modern LLM-driven approaches, highlighting how LLMs contribute unique advantages. Finally, we discuss open research questions and promising future directions in this emerging interdisciplinary field. To support further exploration, we have curated a repository of recent advances on graph learning challenges: https://github.com/limengran98/Awesome-Literature-Graph-Learning-Challenges.

cs.LG↗

arXiv2025

MDVT: Enhancing Multimodal Recommendation with Model-Agnostic Multimodal-Driven Virtual Triplets

The data sparsity problem significantly hinders the performance of recommender systems, as traditional models rely on limited historical interactions to learn user preferences and item properties. While incorporating multimodal information can explicitly represent these preferences and properties, existing works often use it only as side information, failing to fully leverage its potential. In this paper, we propose MDVT, a model-agnostic approach that constructs multimodal-driven virtual triplets to provide valuable supervision signals, effectively mitigating the data sparsity problem in multimodal recommendation systems. To ensure high-quality virtual triplets, we introduce three tailored warm-up threshold strategies: static, dynamic, and hybrid. The static warm-up threshold strategy exhaustively searches for the optimal number of warm-up epochs but is time-consuming and computationally intensive. The dynamic warm-up threshold strategy adjusts the warm-up period based on loss trends, improving efficiency but potentially missing optimal performance. The hybrid strategy combines both, using the dynamic strategy to find the approximate optimal number of warm-up epochs and then refining it with the static strategy in a narrow hyper-parameter space. Once the warm-up threshold is satisfied, the virtual triplets are used for joint model optimization by our enhanced pair-wise loss function without causing significant gradient skew. Extensive experiments on multiple real-world datasets demonstrate that integrating MDVT into advanced multimodal recommendation models effectively alleviates the data sparsity problem and improves recommendation performance, particularly in sparse data scenarios.

cs.IR↗

arXiv2025

MM-STFlowNet: A Transportation Hub-Oriented Multi-Mode Passenger Flow Prediction Method via Spatial-Temporal Dynamic Graph Modeling

Accurate and refined passenger flow prediction is essential for optimizing the collaborative management of multiple collection and distribution modes in large-scale transportation hubs. Traditional methods often focus only on the overall passenger volume, neglecting the interdependence between different modes within the hub. To address this limitation, we propose MM-STFlowNet, a comprehensive multi-mode prediction framework grounded in dynamic spatial-temporal graph modeling. Initially, an integrated temporal feature processing strategy is implemented using signal decomposition and convolution techniques to address data spikes and high volatility. Subsequently, we introduce the Spatial-Temporal Dynamic Graph Convolutional Recurrent Network (STDGCRN) to capture detailed spatial-temporal dependencies across multiple traffic modes, enhanced by an adaptive channel attention mechanism. Finally, the self-attention mechanism is applied to incorporate various external factors, further enhancing prediction accuracy. Experiments on a real-world dataset from Guangzhounan Railway Station in China demonstrate that MM-STFlowNet achieves state-of-the-art performance, particularly during peak periods, providing valuable insight for transportation hub management.

cs.LG↗

arXiv2025

PhenoProfiler: Advancing Phenotypic Learning for Image-based Drug Discovery

In the field of image-based drug discovery, capturing the phenotypic response of cells to various drug treatments and perturbations is a crucial step. However, existing methods require computationally extensive and complex multi-step procedures, which can introduce inefficiencies, limit generalizability, and increase potential errors. To address these challenges, we present PhenoProfiler, an innovative model designed to efficiently and effectively extract morphological representations, enabling the elucidation of phenotypic changes induced by treatments. PhenoProfiler is designed as an end-to-end tool that processes whole-slide multi-channel images directly into low-dimensional quantitative representations, eliminating the extensive computational steps required by existing methods. It also includes a multi-objective learning module to enhance robustness, accuracy, and generalization in morphological representation learning. PhenoProfiler is rigorously evaluated on large-scale publicly available datasets, including over 230,000 whole-slide multi-channel images in end-to-end scenarios and more than 8.42 million single-cell images in non-end-to-end settings. Across these benchmarks, PhenoProfiler consistently outperforms state-of-the-art methods by up to 20%, demonstrating substantial improvements in both accuracy and robustness. Furthermore, PhenoProfiler uses a tailored phenotype correction strategy to emphasize relative phenotypic changes under treatments, facilitating the detection of biologically meaningful signals. UMAP visualizations of treatment profiles demonstrate PhenoProfiler ability to effectively cluster treatments with similar biological annotations, thereby enhancing interpretability. These findings establish PhenoProfiler as a scalable, generalizable, and robust tool for phenotypic learning.

cs.LG↗

arXiv2025

Topology-Driven Attribute Recovery for Attribute Missing Graph Learning in Social Internet of Things

With the advancement of information technology, the Social Internet of Things (SIoT) has fostered the integration of physical devices and social networks, deepening the study of complex interaction patterns. Text Attribute Graphs (TAGs) capture both topological structures and semantic attributes, enhancing the analysis of complex interactions within the SIoT. However, existing graph learning methods are typically designed for complete attributed graphs, and the common issue of missing attributes in Attribute Missing Graphs (AMGs) increases the difficulty of analysis tasks. To address this, we propose the Topology-Driven Attribute Recovery (TDAR) framework, which leverages topological data for AMG learning. TDAR introduces an improved pre-filling method for initial attribute recovery using native graph topology. Additionally, it dynamically adjusts propagation weights and incorporates homogeneity strategies within the embedding space to suit AMGs' unique topological structures, effectively reducing noise during information propagation. Extensive experiments on public datasets demonstrate that TDAR significantly outperforms state-of-the-art methods in attribute reconstruction and downstream tasks, offering a robust solution to the challenges posed by AMGs. The code is available at https://github.com/limengran98/TDAR.

cs.AI↗

arXiv2025

AttriReBoost: A Gradient-Free Propagation Optimization Method for Cold Start Mitigation in Attribute Missing Graphs

Missing attribute issues are prevalent in the graph learning, leading to biased outcomes in Graph Neural Networks (GNNs). Existing methods that rely on feature propagation are prone to cold start problem, particularly when dealing with attribute resetting and low-degree nodes, which hinder effective propagation and convergence. To address these challenges, we propose AttriReBoost (ARB), a novel method that incorporates propagation-based method to mitigate cold start problems in attribute-missing graphs. ARB enhances global feature propagation by redefining initial boundary conditions and strategically integrating virtual edges, thereby improving node connectivity and ensuring more stable and efficient convergence. This method facilitates gradient-free attribute reconstruction with lower computational overhead. The proposed method is theoretically grounded, with its convergence rigorously established. Extensive experiments on several real-world benchmark datasets demonstrate the effectiveness of ARB, achieving an average accuracy improvement of 5.11% over state-of-the-art methods. Additionally, ARB exhibits remarkable computational efficiency, processing a large-scale graph with 2.49 million nodes in just 16 seconds on a single GPU. Our code is available at https://github.com/limengran98/ARB.

cs.LG↗

arXiv2024

TAS-TsC: A Data-Driven Framework for Estimating Time of Arrival Using Temporal-Attribute-Spatial Tri-space Coordination of Truck Trajectories

Accurately estimating time of arrival (ETA) for trucks is crucial for optimizing transportation efficiency in logistics. GPS trajectory data offers valuable information for ETA, but challenges arise due to temporal sparsity, variable sequence lengths, and the interdependencies among multiple trucks. To address these issues, we propose the Temporal-Attribute-Spatial Tri-space Coordination (TAS-TsC) framework, which leverages three feature spaces-temporal, attribute, and spatial-to enhance ETA. Our framework consists of a Temporal Learning Module (TLM) using state space models to capture temporal dependencies, an Attribute Extraction Module (AEM) that transforms sequential features into structured attribute embeddings, and a Spatial Fusion Module (SFM) that models the interactions among multiple trajectories using graph representation learning.These modules collaboratively learn trajectory embeddings, which are then used by a Downstream Prediction Module (DPM) to estimate arrival times. We validate TAS-TsC on real truck trajectory datasets collected from Shenzhen, China, demonstrating its superior performance compared to existing methods.

cs.AI↗

arXiv2024

Effective Finite Time Stability Control for Human-Machine Shared Vehicle Following System

With the development of intelligent connected vehicle technology, human-machine shared control has gained popularity in vehicle following due to its effectiveness in driver assistance. However, traditional vehicle following systems struggle to maintain stability when driver reaction time fluctuates, as these variations require different levels of system intervention. To address this issue, the proposed human-machine shared vehicle following assistance system (HM-VFAS) integrates driver outputs under various states with the assistance system. The system employs an intelligent driver model that accounts for reaction time delays, simulating time-varying driver outputs. A control authority allocation strategy is designed to dynamically adjust the level of intervention based on real-time driver state assessment. To handle instability from driver authority switching, the proposed solution includes a two-layer adaptive finite time sliding mode controller (A-FTSMC). The first layer is an integral sliding mode adaptive controller that ensures robustness by compensating for uncertainties in the driver output. The second layer is a fast non-singular terminal sliding mode controller designed to accelerate convergence for rapid stabilization. Using real driver videos as inputs, the performance of the HM-VFAS was evaluated. Results show that the proposed control strategy maintains a safe distance under time-varying driver states, with the actual acceleration error relative to the target acceleration maintained within 0.5m/s~2 and the maximum acceleration error reduced by 1.2m/s~2. Compared to traditional controllers, the A-FTSMC controller offers faster convergence and less vibration, reducing the stabilization time by 27.3%.

eess.SY↗

arXiv2023

AmGCL: Feature Imputation of Attribute Missing Graph via Self-supervised Contrastive Learning

Attribute graphs are ubiquitous in multimedia applications, and graph representation learning (GRL) has been successful in analyzing attribute graph data. However, incomplete graph data and missing node attributes can have a negative impact on media knowledge discovery. Existing methods for handling attribute missing graph have limited assumptions or fail to capture complex attribute-graph dependencies. To address these challenges, we propose Attribute missing Graph Contrastive Learning (AmGCL), a framework for handling missing node attributes in attribute graph data. AmGCL leverages Dirichlet energy minimization-based feature precoding to encode in missing attributes and a self-supervised Graph Augmentation Contrastive Learning Structure (GACLS) to learn latent variables from the encoded-in data. Specifically, AmGCL utilizies feature reconstruction based on structure-attribute energy minimization while maximizes the lower bound of evidence for latent representation mutual information. Our experimental results on multiple real-world datasets demonstrate that AmGCL outperforms state-of-the-art methods in both feature imputation and node classification tasks, indicating the effectiveness of our proposed method in real-world attribute graph analysis tasks.

cs.LG↗

arXiv2022

Design and testing of LGAD sensor with shallow carbon implantation

The low gain avalanche detectors (LGADs) are thin sensors with fast charge collection which in combination with internal gain deliver an outstanding time resolution of about 30 ps. High collision rates and consequent large particle rates crossing the detectors at the upgraded Large Hadron Collider (LHC) in 2028 will lead to radiation damage and deteriorated performance of the LGADs. The main consequence of radiation damage is loss of gain layer doping (acceptor removal) which requires an increase of bias voltage to compensate for the loss of charge collection efficiency and consequently time resolution. The Institute of High Energy Physics (IHEP), Chinese Academy of Sciences (CAS) has developed a process based on the Institute of Microelectronics (IME), CAS capability to enrich the gain layer with carbon to reduce the acceptor removal effect by radiation. After 1 MeV neutron equivalent fluence of 2.5$\times$10$^{15}$ n$_{eq}$/cm$^{2}$, which is the maximum fluence to which sensors will be exposed at ATLAS High Granularity Timing Detector (HGTD), the IHEP-IME second version (IHEP-IMEv2) 50 $\mu$m LGAD sensors already deliver adequate charge collection > 4 fC and time resolution < 50 ps at voltages < 400 V. The operation voltages of these 50 $\mu$m devices are well below those at which single event burnout may occur.

physics.ins-det↗