Searcharxiv⌕ Search

arXiv subjects

Tao Chen

Publications and source records attributed to Tao Chen.

At least 109 records · Page 6Linked to original sources
arXiv2026

A trust-region funnel algorithm for gray-box optimization

Gray-box optimization, where parts of optimization problems are represented by algebraic models while others are treated as black-box models lacking analytic derivatives, remains a challenge. Trust-region (TR) methods provide a robust framework for gray-box problems through local reduced models (RMs) for black-box components, but they are complex and require extensive parameter tuning. Motivated by recent advances in funnel-based convergence theory for nonlinear optimization, we propose a novel TR funnel algorithm for gray-box optimization, replacing the filter acceptance criterion with a uni-dimensional funnel, maintaining a monotonically decreasing upper bound on approximation error of local black-box RMs. A global convergence proof to a first-order critical point is established. The algorithm, implemented open-source in Pyomo, supports multiple RM forms and globalization strategies (filter or funnel). Benchmark tests show the TR funnel algorithm achieves comparable and often improved performance relative to the classical TR filter method, thus providing a simpler, effective alternative for gray-box optimization.

math.OC↗
arXiv2026

ForestPrune: High-ratio Visual Token Compression for Video Multimodal Large Language Models via Spatial-Temporal Forest Modeling

Due to the great saving of computation and memory overhead, token compression has become a research hot-spot for MLLMs and achieved remarkable progress in image-language tasks. However, for the video, existing methods still fall short of high-ratio token compression. We attribute this shortcoming to the insufficient modeling of temporal and continual video content, and propose a novel and training-free token pruning method for video MLLMs, termed ForestPrune, which achieves effective and high-ratio pruning via Spatial-temporal Forest Modeling. In practice, ForestPrune construct token forests across video frames based on the semantic, spatial and temporal constraints, making an overall comprehension of videos. Afterwards, ForestPrune evaluates the importance of token trees and nodes based on tree depth and node roles, thereby obtaining a globally optimal pruning decision. To validate ForestPrune, we apply it to two representative video MLLMs, namely LLaVA-Video and LLaVA-OneVision, and conduct extensive experiments on a bunch of video benchmarks. The experimental results not only show the great effectiveness for video MLLMs, e.g., retaining 95.8% average accuracy while reducing 90% tokens for LLaVA-OneVision, but also show its superior performance and efficiency than the compared token compression methods, e.g., +10.1% accuracy on MLVU and -81.4% pruning time than FrameFusion on LLaVA-Video.

cs.CV↗
arXiv2026

PaceLLM: Brain-Inspired Large Language Models for Long-Context Understanding

While Large Language Models (LLMs) demonstrate strong performance across domains, their long-context capabilities are limited by transient neural activations causing information decay and unstructured feed-forward network (FFN) weights leading to semantic fragmentation. Inspired by the brain's working memory and cortical modularity, we propose PaceLLM, featuring two innovations: (1) a Persistent Activity (PA) Mechanism that mimics prefrontal cortex (PFC) neurons' persistent firing by introducing an activation-level memory bank to dynamically retrieve, reuse, and update critical FFN states, addressing contextual decay; and (2) Cortical Expert (CE) Clustering that emulates task-adaptive neural specialization to reorganize FFN weights into semantic modules, establishing cross-token dependencies and mitigating fragmentation. Extensive evaluations show that PaceLLM achieves 6% improvement on LongBench's Multi-document QA and 12.5-17.5% performance gains on Infinite-Bench tasks, while extending measurable context length to 200K tokens in Needle-In-A-Haystack (NIAH) tests. This work pioneers brain-inspired LLM optimization and is complementary to other works. Besides, it can be generalized to any model and enhance their long-context performance and interpretability without structural overhauls.

q-bio.NC↗
arXiv2026

Crosstalk-robust superconducting two-qubit geometric gates using tunable couplers

The design of coupler-based superconducting two-qubit gates simplifies circuit layout and alleviate frequency crowding, thereby enhancing the scalability and flexibility of quantum chips. However, in such architectures, a trade-off often exists between suppressing crosstalk and reducing gate duration, and how to achieve synergistic optimization of both remains an open challenge. To address this, this paper proposes a coupler-assisted superconducting two-qubit geometric gate scheme oriented towards crosstalk robustness. By introducing additional parametric degrees of freedom, the scheme steers the system evolution along desired trajectories, thereby flexibly avoiding crosstalk-sensitive operational regions. Numerical simulations demonstrate that the proposed scheme can effectively suppress crosstalk errors while enabling fast gate operations, and exhibits strong robustness against typical experimental imperfections such as qubit frequency drift. Moreover, even when accounting for unavoidable high-frequency oscillation terms and qubit decoherence in realistic physical systems, our crosstalk-robust two-qubit geometric gates still achieve high fidelity. This work provides a feasible pathway toward robust and efficient two-qubit gate implementation in superconducting quantum computation.

quant-ph↗
arXiv2026

Towards Effective Long Video Understanding of Multimodal Large Language Models via One-shot Clip Retrieval

Due to excessive memory overhead, most Multimodal Large Language Models (MLLMs) can only process videos of limited frames. In this paper, we propose an effective and efficient paradigm to remedy this shortcoming, termed One-shot video-Clip based Retrieval-Augmented Generation (OneClip-RAG). Compared with existing video RAG methods, OneClip-RAG makes full use of the merits of video clips for augmented video understanding in terms of both knowledge integrity and semantic coherence. Besides, it is also equipped with a novel query-guided video chunking algorithm that can unify clip chunking and cross-modal retrieval in one processing step, avoiding redundant computations. To improve instruction following, we further propose a new dataset called SynLongVideo and design a progressive training regime for OneClip-RAG. OneClip-RAG is plugged into three recent MLLMs and validated on a set of long-video benchmarks. Experimental results not only show the obvious performance gains by OneClip-RAG over MLLMs, e.g., boosting Qwen3-VL 8B to the level of GPT-5 on MLVU, but also show its superior efficiency in handling long videos. e.g., enabling LLaVA-Video understand up to an hour of videos in less than 1.2 minutes on a single 4090 GPU.

cs.CV↗
arXiv2026

MedXIAOHE: A Comprehensive Recipe for Building Medical MLLMs

We present MedXIAOHE, a medical vision-language foundation model designed to advance general-purpose medical understanding and reasoning in real-world clinical applications. MedXIAOHE achieves state-of-the-art performance across diverse medical benchmarks and surpasses leading closed-source multimodal systems on multiple capabilities. To achieve this, we propose an entity-aware continual pretraining framework that organizes heterogeneous medical corpora to broaden knowledge coverage and reduce long-tail gaps (e.g., rare diseases). For medical expert-level reasoning and interaction, MedXIAOHE incorporates diverse medical reasoning patterns via reinforcement learning and tool-augmented agentic training, enabling multi-step diagnostic reasoning with verifiable decision traces. To improve reliability in real-world use, MedXIAOHE integrates user-preference rubrics, evidence-grounded reasoning, and low-hallucination long-form report generation, with improved adherence to medical instructions. We release this report to document our practical design choices, scaling insights, and evaluation framework, hoping to inspire further research.

cs.CL↗
arXiv2026

Percolation in the three-dimensional Ising model

Geometric representations provide a useful perspective on critical phenomena in the Ising model. In a recent study [Phys. Rev. E 112, 034118 (2025)], we found that the two-dimensional critical Ising model exhibits two consecutive percolation transitions for geometric spin clusters as the bond-occupation probability $p$ between parallel spins increases. Here, through extensive Monte Carlo simulations, we show that this phenomenon does not persist in three dimensions, where we observe only a single percolation transition on critical Ising configurations. Further theoretical analysis of the Ising model on the complete graph also yields the same scenario. In addition, we study percolation on a two-dimensional layer embedded in the three-dimensional critical Ising model. For this layer system, we estimate the red-bond exponent $y_p = 0.426(6)$ and the fractal dimensions of the largest cluster, hull, and shortest path as $d_f = 1.8926(20)$, $d_{\rm hull} = 1.663(4)$, and $d_{\rm min} = 1.080(10)$, respectively. These values indicate a distinct universality class induced by coupling to out-of-plane critical correlations.

cond-mat.stat-mech↗
arXiv2026

Leakage Suppression in Quantum Control via Static Parameter Offsets

High-fidelity quantum operations require the system dynamics to be strictly confined to the computational subspace. In practice, however, control fields inevitably couple to leakage levels, giving rise to quantum state leakage that significantly reduces the fidelity of the operation. To address this challenge, we propose a general strategy for actively suppressing leakage errors by applying small, static offsets to tunable system parameters. This approach systematically mitigates leakage's detrimental impact on quantum control, without modifying the original control framework or incurring additional time overhead. By avoiding the need for extra suppression pulses or complex optimization procedures altogether, it offers a streamlined solution for leakage compensation while remaining fully compatible with subsequent optimal control techniques. Numerical validation conducted on superconducting quantum circuits demonstrates effective leakage suppression, enabling high-fidelity single-qubit gates, precise control of two-qubit interactions, and perfect state transfer in multi-level systems. Moreover, when integrated with optimal control techniques, our approach also allows for the cooperative suppression of both leakage errors and residual crosstalk. Therefore, this work provides a feasible technical pathway toward the low error thresholds required for fault-tolerant quantum computation.

quant-ph↗
arXiv2026

Universal Robust Quantum Gates via Doubly Geometric Control

Geometric quantum computation offers a potential route to fault-tolerant quantum information processing by exploiting the global nature of geometric phases. However, achieving controlled high-order suppression of multiple error sources remains a long-standing limitation, particularly in realistic large-scale circuits with complex noise environments. This limitation is largely due to the absence of a general framework that directly characterizes error accumulation and enables systematic improvement. Here we establish such a framework for universal doubly geometric gates by embedding target operations into a hierarchy of level-n identity constructions. This approach enables direct quantification of error accumulation while removing structural constraints inherent in previous schemes. We analytically show that the defining conditions lead to simultaneous fourth-order suppression of control errors, with a systematic extension to sixth-order suppression via higher-level constructions. Our results establish doubly geometric control as a general and scalable route toward high-order robust quantum gates, with potential implications for fault-tolerant quantum information processing.

quant-ph↗
arXiv2026

Scaling the Long Video Understanding of Multimodal Large Language Models via Visual Memory Mechanism

Long video understanding is a key challenge that plagues the advancement of \emph{Multimodal Large language Models} (MLLMs). In this paper, we study this problem from the perspective of visual memory mechanism, and proposed a novel and training-free approach, termed \emph{Flexible Memory} (\textbf{FlexMem}). In principle, FlexMem aims to mimic human behavior of video watching, \emph{i.e.}, continually watching video content and recalling the most relevant memory fragments to answer the question. In this way, FlexMem can help MLLMs achieve video understanding of infinite lengths, unlike previous methods that process all video information at once and have input upper-limit. Concretely, FlexMem first consider the visual KV caches as the memory sources, and realize the effective memory transfer and writing via a dual-pathway compression design. Afterwards, FlexMem also explores different memory reading strategies for the diverse video understanding tasks, including the popular streaming one. To validate FlexMem, we apply it to two popular video-MLLMs, and conduct extensive experiments on five long video and one streaming video task. The experimental results show that on \textbf{a single 3090 GPU}, our FlexMem can achieve obvious improvements than existing efficient video understanding methods and process more than \textbf{1k frames}, which also helps the base MLLMs achieve comparable or even better performance than SOTA MLLMs on some benchmarks, \emph{e.g.} , GPT-4o and Gemini-1.5 Pro.

cs.CV↗
arXiv2026

Simultaneous measurement of pressure-dependent bulk and interfacial thermal properties in thermal interface materials using square-pulsed source thermoreflectance

Thermal interface materials (TIMs) critically regulate heat dissipation from electronic chips to heat spreaders, yet their thermal conductivity (k), volumetric heat capacity (C), and interfacial thermal resistance (ITR) evolve with mechanical pressure and cannot be determined simultaneously using existing steady-state or transient techniques. As a result, the coupled roles of bulk compaction and interfacial contact in governing heat transport in TIM assemblies remain poorly resolved. Here, we present a square-pulsed source (SPS) thermoreflectance method that enables simultaneous determination of k, C, and ITR in TIM stacks under controlled mechanical loading. By spanning square-wave modulation frequencies from 1 Hz to 10 MHz, SPS probes a broad range of thermal penetration depths, enabling distinction between heat diffusion in the TIM bulk and interfacial heat transfer at the Al/TIM contact. Measurements on a thermally conductive gel, a thermal pad, and a high-vacuum grease during compression-unloading cycles reveal distinct pressure-dependent thermal transport mechanisms. The gel and pad exhibit increases in k and C, reduced ITR, and pronounced hysteresis, indicating coupled bulk densification and persistent interfacial conformity during loading cycles. In contrast, the grease shows nearly pressure-independent bulk properties but a strong pressure dependence of ITR, consistent with an interface-dominated response. These results resolve the long-standing challenge of simultaneously quantifying bulk and interfacial thermal transport in mechanically loaded TIM assemblies, enabling experimentally constrained thermal management and reliability analysis in electronic packaging.

physics.app-ph↗
arXiv2026

A Nonlinear Incremental Approach for Replay Attack Detection

Replay attacks comprise replaying previously recorded sensor measurements and injecting malicious signals into a physical plant, causing great damage to cyber-physical systems. Replay attack detection has been widely studied for linear systems, whereas limited research has been reported for nonlinear cases. In this paper, the replay attack is studied in the context of a nonlinear plant controlled by an observer-based output feedback controller. We first analyze replay attack detection using an innovation-based detector and reveal that this detector alone may fail to detect such attacks. Consequently, we turn to a watermark-based design framework to improve the detection. In the proposed framework, the effects of the watermark on attack detection and closed-loop system performance loss are quantified by two indices, which exploit the incremental gains of nonlinear systems. To balance the detection performance and control system performance loss, an explicit optimization problem is formulated. Moreover, to achieve a better balance, we generalize the proposed watermark design framework to co-design the watermark, controller and observer. Numerical simulations are presented to validate the proposed frameworks.

eess.SY↗
arXiv2026

Weak Coupling of Diffusional and Phonon-like Modes in Liquids Revealed by Dynamic Kapitza Length

Understanding heat transfer across solid-liquid interfaces is central to thermal management and energy technologies, yet whether the interfacial thermal conductance (ITC) depends on the timescale of heating remains unclear. Here we use square-pulsed source thermoreflectance, which combines time-resolved detection with broadband modulation, to probe Al-water and Al-octane interfaces. We observe a reproducible increase of the apparent ITC with modulation frequency. A control Al-silica interface shows no measurable frequency dependence, indicating that the effect is specific to liquids rather than a generic feature shared by all amorphous materials. We explain the data with a two-channel liquid picture in which diffusional and phonon-like modes exchange energy weakly over a finite nonequilibrium length. From the relative magnitude of the thermal penetration depth and the nonequilibrium length, we identify three transport regimes. These findings challenge the common assumption of fully equilibrated liquid modes and provide experimental constraints for modeling dynamic energy exchange at liquid interfaces.

cond-mat.mtrl-sci↗
arXiv2026

PEARL: Geometry Aligns Semantics for Training-Free Open-Vocabulary Semantic Segmentation

Training-free open-vocabulary semantic segmentation (OVSS) promises rapid adaptation to new label sets without retraining. Yet, many methods rely on heavy post-processing or handle text and vision in isolation, leaving cross-modal geometry underutilized. Others introduce auxiliary vision backbones or multi-model pipelines, which increase complexity and latency while compromising design simplicity. We present PEARL, \textbf{\underline{P}}rocrust\textbf{\underline{e}}s \textbf{\underline{a}}lignment with text-awa\textbf{\underline{r}}e \textbf{\underline{L}}aplacian propagation, a compact two-step inference that follows an align-then-propagate principle. The Procrustes alignment step performs an orthogonal projection inside the last self-attention block, rotating keys toward the query subspace via a stable polar iteration. The text-aware Laplacian propagation then refines per-pixel logits on a small grid through a confidence-weighted, text-guided graph solve: text provides both a data-trust signal and neighbor gating, while image gradients preserve boundaries. In this work, our method is fully training-free, plug-and-play, and uses only fixed constants, adding minimal latency with a small per-head projection and a few conjugate-gradient steps. Our approach, PEARL, sets a new state-of-the-art in training-free OVSS without extra data or auxiliary backbones across standard benchmarks, achieving superior performance under both with-background and without-background protocols.

cs.CV↗
arXiv2026

Revealing Domain-Spatiality Patterns for Configuration Tuning: Domain Knowledge Meets Fitness Landscapes

Configuration tuning for better performance is crucial in quality assurance. Yet, there has long been a mystery on tuners' effectiveness, due to the black-box nature of configurable systems. Prior efforts predominantly adopt static domain analysis (e.g., static taint analysis), which often lacks generalizability, or dynamic data analysis (e.g., benchmarking performance analysis), limiting explainability. In this work, we embrace Fitness Landscape Analysis (FLA) as a bridge between domain knowledge and difficulty of the tuning. We propose Domland, a two-pronged methodology that synergizes the spatial information obtained from FLA and domain-driven analysis to systematically capture the hidden characteristics of configuration tuning cases, explaining how and why a tuner might succeed or fail. This helps to better interpret and contextualize the behavior of tuners and inform tuner design. To evaluate Domland, we conduct a case study of nine software systems and 93 workloads, from which we reveal several key findings: (1) configuration landscapes are inherently system-specific, with no single domain factor (e.g., system area, programming language, or resource intensity) consistently shaping their structure; (2) the core options (e.g., pic-struct of x264), which control the main functional flows, exert a stronger influence on landscape ruggedness (i.e. the difficulty of tuning) compared to resource options (e.g., cpu-independent of x264); (3) Workload effects on landscape structure are not uniformly tied to type or scale. Both contribute to landscape variations, but their impact is system-dependent.

cs.SE↗
arXiv2026

Efficiency Follows Global-Local Decoupling

Modern vision models must capture image-level context without sacrificing local detail while remaining computationally affordable. We revisit this tradeoff and advance a simple principle: decouple the roles of global reasoning and local representation. To operationalize this principle, we introduce ConvNeur, a two-branch architecture in which a lightweight neural memory branch aggregates global context on a compact set of tokens, and a locality-preserving branch extracts fine structure. A learned gate lets global cues modulate local features without entangling their objectives. This separation yields subquadratic scaling with image size, retains inductive priors associated with local processing, and reduces overhead relative to fully global attention. On standard classification, detection, and segmentation benchmarks, ConvNeur matches or surpasses comparable alternatives at similar or lower compute and offers favorable accuracy versus latency trade-offs at similar budgets. These results support the view that efficiency follows global-local decoupling.

cs.CV↗
arXiv2026

CurveStream: Boosting Streaming Video Understanding in MLLMs via Curvature-Aware Hierarchical Visual Memory Management

Multimodal Large Language Models have achieved significant success in offline video understanding, yet their application to streaming videos is severely limited by the linear explosion of visual tokens, which often leads to Out-of-Memory (OOM) errors or catastrophic forgetting. Existing visual retention and memory management methods typically rely on uniform sampling, low-level physical metrics, or passive cache eviction. However, these strategies often lack intrinsic semantic awareness, potentially disrupting contextual coherence and blurring transient yet critical semantic transitions. To address these limitations, we propose CurveStream, a training-free, curvature-aware hierarchical visual memory management framework. Our approach is motivated by the key observation that high-curvature regions along continuous feature trajectories closely align with critical global semantic transitions. Based on this geometric insight, CurveStream evaluates real-time semantic intensity via a Curvature Score and integrates an online K-Sigma dynamic threshold to adaptively route frames into clear and fuzzy memory states under a strict token budget. Evaluations across diverse temporal scales confirm that this lightweight framework, CurveStream, consistently yields absolute performance gains of over 10% (e.g., 10.69% on StreamingBench and 13.58% on OVOBench) over respective baselines, establishing new state-of-the-art results for streaming video perception.The code will be released at https://github.com/streamingvideos/CurveStream.

cs.CV↗
arXiv2026

Beyond Quadratic: Linear-Time Change Detection with RWKV

Existing paradigms for remote sensing change detection are caught in a trade-off: CNNs excel at efficiency but lack global context, while Transformers capture long-range dependencies at a prohibitive computational cost. This paper introduces ChangeRWKV, a new architecture that reconciles this conflict. By building upon the Receptance Weighted Key Value (RWKV) framework, our ChangeRWKV uniquely combines the parallelizable training of Transformers with the linear-time inference of RNNs. Our approach core features two key innovations: a hierarchical RWKV encoder that builds multi-resolution feature representation, and a novel Spatial-Temporal Fusion Module (STFM) engineered to resolve spatial misalignments across scales while distilling fine-grained temporal discrepancies. ChangeRWKV not only achieves state-of-the-art performance on the LEVIR-CD benchmark, with an 85.46% IoU and 92.16% F1 score, but does so while drastically reducing parameters and FLOPs compared to previous leading methods. This work demonstrates a new, efficient, and powerful paradigm for operational-scale change detection. Our code and model are publicly available.

cs.CV↗