arXiv subjects

Tao Zou

Publications and source records attributed to Tao Zou.

At least 19 recordsLinked to original sources

arXiv2026

CORE-MTL: Rethinking Gradient Balancing via Causal Orthogonal Representations

Multi-task learning (MTL) aims to construct a joint model for multiple tasks by sharing a common representation across domains. To achieve this goal, existing optimization-centric methods either balance task gradients or modify the shared architecture. However, as these approaches remain agnostic to the content of the shared representation, they fail to disentangle task-relevant structure from spurious context, leading to negative transfer and poor generalization. To overcome this limitation, we propose Causal Orthogonal Representations for Multi-Task Learning (CORE-MTL), a causally motivated representation-centric framework that encourages a structured semantic-residual factorization of the shared representation, concentrating task-relevant structure in the semantic stream while relegating nuisance variation to the residual stream. We instantiate this framework in the visual domain by leveraging physical priors for structured scenes and statistical constraints for attributes. Theoretically, our method enjoys a tighter out-of-distribution generalization bound than optimization-centric methods and reduces task gradient interference without explicit gradient projection or reweighting. Empirically, CORE-MTL consistently outperforms existing methods on visual multi-task benchmarks in both in-distribution and out-of-distribution settings. Code is publicly available at https://github.com/Hope-Rita/CORE-MTL.

cs.CV↗

arXiv2026

Task-Focused Memorization for Multimodal Agents

Long-term memory is essential for multimodal agents to build coherent experience, accumulate world knowledge, and achieve continual learning. However, constructing effective memory goes beyond memory module design and basic requirements such as accuracy and fidelity; the key challenge lies in determining what to memorize. Multimodal agents, such as embodied agents, continuously perceive, reason, and act in real or virtual environments, receiving an unbounded stream of multimodal observations. From this combinatorial explosion of information, an agent must selectively retain content that is relevant to its role in the environment and valuable for future tasks. To bridge this gap, we frame memory generation as a learnable memorization policy and introduce TaskMem (Task-focused Memorization Policy Learning), a reinforcement-learning-based framework that enables the policy to dynamically adjust its focus to the demands of real tasks encountered in the environment. TaskMem adopts a two-phase training paradigm: Phase One learns how to memorize by optimizing memory quality under fundamental fidelity requirements; Phase Two occurs after deployment, where the agent learns what to memorize by tuning an adapter on its base MLLM, using recent environment tasks to define a reward model that guides the memorization policy toward task-relevant content. To evaluate our approach, we reformulate VideoMME, EgoLife, and EgoTempo into streaming benchmarks that simulate a realistic setting in which an agent processes streaming observations and handles tasks arriving online. To isolate memory assessment, the questions must be answered using only the agent's memory, without access to raw video. Built on Qwen3-VL-30B-A3B, TaskMem improves VQA accuracy by 6.3%, 7.0%, and 5.3% on these benchmarks, respectively.

cs.CV↗

arXiv2026

DuConTE: Dual-Granularity Text Encoder with Topology-Constrained Attention for Text-attributed Graphs

Text-attributed graphs integrate semantic information of node texts with topological structure, offering significant value in various applications such as document classification and information extraction. Existing approaches typically encode textual content using language models (LMs), followed by graph neural networks (GNNs) to process structural information. However, during the LM-based text encoding phase, most methods not only perform semantic interaction solely at the word-token granularity, but also neglect the structural dependencies among texts from different nodes. In this work, we propose DuConTE, a dual-granularity text encoder with topology-constrained attention. The model employs a cascaded architecture of two pretrained LMs, encoding semantics first at the word-token granularity and then at the node granularity. During the self-attention computation in each LM, we dynamically adjust the attention mask matrix based on node connectivity, guiding the model to learn semantic correlations informed by the graph structure. Furthermore, when composing node representations from word-token embeddings, we separately evaluate the importance of tokens under the center-node context and the neighborhood context, enabling the capture of more contextually relevant semantic information. Extensive experiments on multiple benchmark datasets demonstrate that DuConTE achieves state-of-the-art performance on the majority of them.

cs.CL↗

arXiv2026

A Clinical Point Cloud Paradigm for In-Hospital Mortality Prediction from Multi-Level Incomplete Multimodal EHRs

Deep learning-based modeling of multimodal Electronic Health Records (EHRs) has become an important approach for clinical diagnosis and risk prediction. However, due to diverse clinical workflows and privacy constraints, raw EHRs are inherently multi-level incomplete, including irregular sampling, missing modalities, and sparse labels. These issues cause temporal misalignment, modality imbalance, and limited supervision. Most existing multimodal methods assume relatively complete data, and even methods designed for incompleteness usually address only one or two of these issues in isolation. As a result, they often rely on rigid temporal/modal alignment or discard incomplete data, which may distort raw clinical semantics. To address this problem, we propose HealthPoint (HP), a unified clinical point cloud paradigm for multi-level incomplete EHRs. HP represents heterogeneous clinical events as points in a continuous 4D space defined by content, time, modality, and case. To model interactions between arbitrary point pairs, we introduce a Low-Rank Relational Attention mechanism that efficiently captures high-order dependencies across these four dimensions. We further develop a hierarchical interaction and sampling strategy to balance fine-grained modeling and computational efficiency. Built on this framework, HP enables flexible event-level interaction and fine-grained self-supervision, supporting robust modality recovery and effective use of unlabeled data. Experiments on large-scale EHR datasets for risk prediction show that HP consistently achieves state-of-the-art performance and strong robustness under varying degrees of incompleteness.

cs.LG↗

arXiv2026

Hardware implementation of photonic neuromorphic autonomous navigation

Reinforcement learning (RL) is a core technology enabling the transition of artificial intelligence (AI) from perception to decision-making, but its deployment on conventional electronic hardware suffers from high latency and energy consumption imposed by the von Neumann architecture. Here, we propose a photonic spiking twin delayed deep deterministic policy gradient (TD3) reinforcement learning architecture for neuromorphic autonomous navigation and experimentally validate it on a distributed feedback laser with a saturable absorber (DFB-SA) array. The hybrid architecture integrates a photonic spiking Actor network with dual continuous-valued Critic networks, where the final nonlinear spiking activation layer of the Actor is deployed on the DFB-SA laser array. In autonomous navigation tasks, the system achieves an average reward of 58.22 plus-minus 17.29 and a success rate of 80% plus-minus 8.3%. Hardware-software co-inference demonstrates an estimated energy consumption of 0.78 nJ/inf and an ultra-low latency of 191.20 ps/inf, with co-inference error rates of 0.051% and 0.059% in task scenarios with and without obstacle interference, respectively. Simulations for error-activated channels show full agreement with the expected responses, validating the dynamic characteristics of the DFB-SA laser. The architecture shows strong potential for integration with large-scale photonic linear computing chips, enabling fully-functional photonic computation and low-power, low-latency neuromorphic autonomous navigation.

physics.optics↗

arXiv2026

Incident-Guided Spatiotemporal Traffic Forecasting

Recent years have witnessed the rapid development of deep-learning-based, graph-neural-network-based forecasting methods for modern intelligent transportation systems. However, most existing work focuses exclusively on capturing spatio-temporal dependencies from historical traffic data, while overlooking the fact that suddenly occurring transportation incidents, such as traffic accidents and adverse weather, serve as external disturbances that can substantially alter temporal patterns. We argue that this issue has become a major obstacle to modeling the dynamics of traffic systems and improving prediction accuracy, but the unpredictability of incidents makes it difficult to observe patterns from historical sequences. To address these challenges, this paper proposes a novel framework named the Incident-Guided Spatiotemporal Graph Neural Network (IGSTGNN). IGSTGNN explicitly models the incident's impact through two core components: an Incident-Context Spatial Fusion (ICSF) module to capture the initial heterogeneous spatial influence, and a Temporal Incident Impact Decay (TIID) module to model the subsequent dynamic dissipation. To facilitate research on the spatio-temporal impact of incidents on traffic flow, a large-scale dataset is constructed and released, featuring incident records that are time-aligned with traffic time series. On this new benchmark, the proposed IGSTGNN framework is demonstrated to achieve state-of-the-art performance. Furthermore, the generalizability of the ICSF and TIID modules is validated by integrating them into various existing models.

cs.LG↗

arXiv2026

DisCo-FLoc: Semantic-Free Floorplan Localization via $SE(2)$-Aware Contrastive Disambiguation

Visual Floorplan Localization (FLoc) struggles with severe structural aliasing caused by repetitive minimalist layouts. This occurs because physically distant poses share highly similar visual-geometric features, which degrades spatial separability and angular discriminability. While existing methods attempt to mitigate these ambiguities by relying on costly semantic annotations, the resulting performance gains remain inherently limited. To address the above issues, we propose DisCo-FLoc, a semantic-free method for visual-geometric Contrastive Disambiguation. First, we introduce a depth-aware Ray Regression Predictor (RRP) that serves as a dense-to-ray geometric projector. By explicitly suppressing visual clutter along the vertical dimension, RRP projects monocular RGB images into 2D ray primitives, which are matched with floorplans to produce geometry-aware FLoc candidates. Second, to resolve the remaining ambiguity among these candidates, we propose a spatially perturbed contrastive objective to align RGB images with local floorplan structures and formulate a visual-geometric compatibility function. In particular, we meticulously construct positive and negative samples at both positional and directional levels through $SE(2)$ pose perturbations for contrastive learning, effectively achieving pose smoothness, spatial separability, and angular discriminability. The compatibility function enables DisCo-FLoc to disambiguate FLoc by using richer visual context beyond pure geometric layouts, without requiring any semantic annotations. Extensive experiments on two challenging visual FLoc benchmarks demonstrate that DisCo-FLoc significantly outperforms state-of-the-art semantic-based methods, especially narrowing the performance gap between positional and directional FLoc accuracy.

cs.RO↗

arXiv2025

Guidewire-driven deployment of high density ECoG arrays for large area brain-computer interface

Electrocorticographic brain computer interfaces are powerful emergent technologies for advancing basic neuroscience research and targeted clinical interventions. However, existing devices require trade-offs between coverage area, electrode density, surgical invasiveness and complication risk: limitations that fail to meet the demands of next-generation BCI. Here, we report a guidewire-driven deployable ECoG BCI device that can be epidurally implanted using minimally invasive procedures. Our ultra-flexible but strong thin-film electrode array, which packs 256 electrodes into 4 cm2, can be folded, pulled through millimetre-sized skull holes, and unfurled seamlessly onto the brain dura mater. When deployed on the canine brain, it captures abundant high-quality auditory neural signals with distinct features of hearing that can be used to classify sound types with over 80% accuracy using various standard machine learning models. Our device is biocompatible for chronic monitoring, easy and fast to deploy and importantly, resolves the key trade-offs limiting current BCI technologies.

physics.med-ph↗

arXiv2025

Global-Lens Transformers: Adaptive Token Mixing for Dynamic Link Prediction

Dynamic graph learning plays a pivotal role in modeling evolving relationships over time, especially for temporal link prediction tasks in domains such as traffic systems, social networks, and recommendation platforms. While Transformer-based models have demonstrated strong performance by capturing long-range temporal dependencies, their reliance on self-attention results in quadratic complexity with respect to sequence length, limiting scalability on high-frequency or large-scale graphs. In this work, we revisit the necessity of self-attention in dynamic graph modeling. Inspired by recent findings that attribute the success of Transformers more to their architectural design than attention itself, we propose GLFormer, a novel attention-free Transformer-style framework for dynamic graphs. GLFormer introduces an adaptive token mixer that performs context-aware local aggregation based on interaction order and time intervals. To capture long-term dependencies, we further design a hierarchical aggregation module that expands the temporal receptive field by stacking local token mixers across layers. Experiments on six widely-used dynamic graph benchmarks show that GLFormer achieves SOTA performance, which reveals that attention-free architectures can match or surpass Transformer baselines in dynamic graph settings with significantly improved efficiency.

cs.LG↗

arXiv2025

Quantifying Periodicity in Non-Euclidean Object-Valued Time Series

Non-Euclidean object-valued time series are playing a growing role in modern data analysis, and periodicity is a fundamental characteristic of many time series. However, quantifying periodicity in general non-Euclidean random objects remains largely unexplored. In this work, we introduce a novel nonparametric framework for quantifying periodicity in random objects within a general metric space that lacks Euclidean structure. Our approach formulates periodicity estimation as a model selection problem and provides methodologies for period estimation, data-driven tuning parameter selection, and periodic component extraction. Our theoretical contributions include deriving uniform convergence rates for general Fr\'echet regression under temporal dependence, establishing the consistency of period estimation without relying on linearity properties used in the literature for Euclidean data, providing theoretical support for data-driven tuning parameter selection, and deriving uniform convergence results for periodic component estimation. Through extensive simulation studies covering three distinct types of time-varying random objects such as compositional data, networks, and functional data, we showcase the excellent accuracy achieved by our approach in periodicity quantification. Finally, we apply our method to various real datasets, including compositional data arising in U.S. electricity generation, New York City transportation networks, and Germany's water consumption curves, highlighting its practical relevance in identifying and quantifying meaningful periodic patterns.

stat.ME↗

arXiv2025

EIFBENCH: Extremely Complex Instruction Following Benchmark for Large Language Models

With the development and widespread application of large language models (LLMs), the new paradigm of "Model as Product" is rapidly evolving, and demands higher capabilities to address complex user needs, often requiring precise workflow execution which involves the accurate understanding of multiple tasks. However, existing benchmarks focusing on single-task environments with limited constraints lack the complexity required to fully reflect real-world scenarios. To bridge this gap, we present the Extremely Complex Instruction Following Benchmark (EIFBENCH), meticulously crafted to facilitate a more realistic and robust evaluation of LLMs. EIFBENCH not only includes multi-task scenarios that enable comprehensive assessment across diverse task types concurrently, but also integrates a variety of constraints, replicating complex operational environments. Furthermore, we propose the Segment Policy Optimization (SegPO) algorithm to enhance the LLM's ability to accurately fulfill multi-task workflow. Evaluations on EIFBENCH have unveiled considerable performance discrepancies in existing LLMs when challenged with these extremely complex instructions. This finding underscores the necessity for ongoing optimization to navigate the intricate challenges posed by LLM applications.

cs.CL↗

arXiv2025

PRIME: Physics-Related Intelligent Mixture of Experts for Transistor Characteristics Prediction

In recent years, machine learning has been extensively applied to data prediction during process ramp-up, with a particular focus on transistor characteristics for circuit design and manufacture. However, capturing the nonlinear current response across multiple operating regions remains a challenge for neural networks. To address such challenge, a novel machine learning framework, PRIME (Physics-Related Intelligent Mixture of Experts), is proposed to capture and integrate complex regional characteristics. In essence, our framework incorporates physics-based knowledge with data-driven intelligence. By leveraging a dynamic weighting mechanism in its gating network, PRIME adaptively activates the suitable expert model based on distinct input data features. Extensive evaluations are conducted on various gate-all-around (GAA) structures to examine the effectiveness of PRIME and considerable improvements (60\%-84\%) in prediction accuracy are shown over state-of-the-art models.

cs.LG↗

arXiv2025

Regularization and Selection in A Directed Network Model with Nodal Homophily and Nodal Effects

This article introduces a regularization and selection methods for directed networks with nodal homophily and nodal effects. The proposed approach not only preserves the statistical efficiency of the resulting estimator, but also ensures that the selection of nodal homophily and nodal effects is scalable with large-scale network data and multiple nodal features. In particular, we propose a directed random network model with nodal homophily and nodal effects, which includes the nodal features in the probability density of random networks. Subsequently, we propose a regularized maximum likelihood estimator with an adaptive LASSO-type regularizer. We demonstrate that the regularized estimator exhibits the consistency and possesses the oracle properties. In addition, we propose a network Bayesian information criterion which ensures the selection consistency while tuning the model. Simulation experiments are conducted to demonstrate the excellent numerical performance. An online friendship network among musicians with nodal musical preference is used to illustrate the usefulness of the proposed new network model in network-related empirical analysis.

stat.ME↗

arXiv2025

Subbagging Variable Selection for Big Data

This article introduces a subbagging (subsample aggregating) approach for variable selection in regression within the context of big data. The proposed subbagging approach not only ensures that variable selection is scalable given the constraints of available computational resources, but also preserves the statistical efficiency of the resulting estimator. In particular, we propose a subbagging loss function that aggregates the least-squares approximations of the loss function for each subsample. Subsequently, we penalize the subbagging loss function via an adaptive LASSO-type regularizer, and obtain a regularized estimator to achieve variable selection. We then demonstrate that the regularized estimator exhibits $\sqrt{N}$-consistency and possesses the oracle properties, where $N$ represents the size of the full sample in the big data. In addition, we propose a subbagging Bayesian information criterion to select the regularization parameter, ensuring that the regularized estimator achieves selection consistency. Simulation experiments are conducted to demonstrate the numerical performance. A U.S. census dataset is analyzed to illustrate the usefulness and computational scalability of the subbagging variable selection method.

stat.ME↗

arXiv2025

On Robust Aggregation for Distributed Data

When data are stored across multiple locations, directly pooling all the data together for statistical analysis may be impossible due to communication costs and privacy concerns. Distributed computing systems allow the analysis of such data, by getting local servers to separately process their own statistical analyses and using a central processor to aggregate the local statistical results. Naive aggregation of local statistics using simple or weighted averages, is vulnerable to contamination within a distributed computing system. This paper develops and investigates a Huber-type aggregation method for locally computed M-estimators to handle contamination in the local estimates. Our implementation of this aggregation method requires estimating the asymptotic variance-covariance matrix of the M-estimator, which we accomplish using a robust spatial median approach. Theoretically, the Huber-type aggregation achieves the same convergence rate as if all the data were pooled. We establish its asymptotic normality for making inferences, including justifying a two-step approach for detecting contamination in the distributed computing system. Extensive simulation studies are conducted to validate the theoretical results and the usefulness of our proposed approach is demonstrated on U.S. airline data.

stat.ME↗

arXiv2025

Fixed and Random Covariance Regression Analyses

Covariance regression analysis is an approach to linking the covariance of responses to a set of explanatory variables $X$, where $X$ can be a vector, matrix, or tensor. Most of the literature on this topic focuses on the "Fixed-$X$" setting and treats $X$ as nonrandom. By contrast, treating explanatory variables $X$ as random, namely the "Random-$X$" setting, is often more realistic in practice. This article aims to fill this gap in the literature on the estimation and model assessment theory for Random-$X$ covariance regression models. Specifically, we construct a new theoretical framework for studying the covariance estimators under the Random-$X$ setting, and we demonstrate that the quasi-maximum likelihood estimator and the weighted least squares estimator are both consistent and asymptotically normal. In addition, we develop pioneering work on the model assessment theory of covariance regression. In particular, we obtain the bias-variance decompositions for the expected test errors under both the Fixed-$X$ and Random-$X$ settings. We show that moving from a Fixed-$X$ to a Random-$X$ setting can increase both the bias and the variance in expected test errors. Subsequently, we propose estimators of the expected test errors under the Fixed-$X$ and Random-$X$ settings, which can be used to assess the performance of the competing covariance regression models. The proposed estimation and model assessment approaches are illustrated via extensive simulation experiments and an empirical study of stock returns in the US market.

math.ST↗

arXiv2025

Change Point Detection for Random Objects with Periodic Behavior

Time-varying random objects have been increasingly encountered in modern data analysis. Moreover, in a substantial number of these applications, periodic behaviour of the random objects has been observed. We develop a novel procedure to identify and localize abrupt changes in the distribution of non-Euclidean random objects with periodic behaviour. The proposed procedure is flexible and broadly applicable, accommodating a variety of suitable change point detectors for random objects. We further construct a specific detector used in the proposed procedure which is nonparametric and effectively captures the entire distribution of these random objects. The theoretical results cover the limiting distribution of the detector under the null hypothesis of no change point, the power of the test in the presence of change points under local alternatives and the consistency in estimating the number and locations of change points, whether dealing with a single change point or multiple ones. We demonstrate that the most competitive method currently in the literature for change point detection in random objects is degraded by periodic behaviour, as periodicity leads to blurring of the changes that this procedure aims to discover. Through comprehensive simulation studies, we demonstrate the superior power and accuracy of our approach in both detecting change points and pinpointing their locations. Our main application is to weighted networks, represented through graph Laplacians. The proposed method delivers highly interpretable results, as evidenced by the identification of meaningful change points in the New York City Citi Bike sharing system that align with significant historical events.

stat.ME↗

arXiv2024

Penalized Sparse Covariance Regression with High Dimensional Covariates

Covariance regression offers an effective way to model the large covariance matrix with the auxiliary similarity matrices. In this work, we propose a sparse covariance regression (SCR) approach to handle the potentially high-dimensional predictors (i.e., similarity matrices). Specifically, we use the penalization method to identify the informative predictors and estimate their associated coefficients simultaneously. We first investigate the Lasso estimator and subsequently consider the folded concave penalized estimation methods (e.g., SCAD and MCP). However, the theoretical analysis of the existing penalization methods is primarily based on i.i.d. data, which is not directly applicable to our scenario. To address this difficulty, we establish the non-asymptotic error bounds by exploiting the spectral properties of the covariance matrix and similarity matrices. Then, we derive the estimation error bound for the Lasso estimator and establish the desirable oracle property of the folded concave penalized estimator. Extensive simulation studies are conducted to corroborate our theoretical results. We also illustrate the usefulness of the proposed method by applying it to a Chinese stock market dataset.

stat.ME↗