Searcharxiv⌕ Search

arXiv subjects

Zhiyuan Zhao

Publications and source records attributed to Zhiyuan Zhao.

At least 19 recordsLinked to original sources
arXiv2026

Nonequilibrium orientational bistability of a magnetically-driven slender particle

Externally driven systems are non-equilibrium but accessible to stationary states. By combining theoretical calculations and particle-based simulations, we investigate the orientational configuration of a transversely-magnetized slender particle in a rotating magnetic field. Our results reveal nonequilibrium orientational bistability in both regimes that the particle rotates synchronously and asynchronously with the field. The relative populations of standing and lying orientations depend strongly on the driving frequency in the synchronous regime, whereas they are primarily controlled by the magnetic field strength in the asynchronous regime. Effective free-energy approximations further capture the principal features of the steady-state orientational distributions in both regimes. Our work aims to give a statistical-thermodynamic description of orientational bistability in magnetically-driven slender particles, paving the way for relevant experimental designs and advanced applications.

cond-mat.soft↗
arXiv2026

Not Every Rollout Needs Distillation: Correctness-Filtered On-Policy Self-Distillation

On-policy self-distillation (OPSD) supervises student-generated trajectories with a reference-conditioned version of the same base model. We ask whether every trajectory supplies distinct supervision that warrants its distillation cost. Gradient measurements on Qwen3-1.7B at initialization reveal that correct and incorrect responses have similar individual gradient magnitudes, but incorrect responses have approximately three times the coherent gradient share. Their estimated population-mean directions also largely overlap. A signal--noise model fitted at a small group size tracks alignment at larger group sizes without refitting. Motivated by these measurements, we introduce correctness-filtered OPSD (C-OPSD), which generates one response per problem and distills only a subset of incorrect responses. Removing unselected rows before both distillation forwards avoids their teacher/student forward and student backward work. Across Qwen3-1.7B, 4B, and 8B on AIME24, AIME25, and HMMT25, C-OPSD achieves on-par performance with full OPSD. In a measured micro-batch-size-four implementation, executed rows decrease by approximately 68%, and with retained rows packed into full micro-batches, training takes about 40% less time while rollout generation remains unchanged. Together, the measurements and training results connect outcome-conditioned supervision redundancy to more compute-efficient distillation.

cs.AI↗
arXiv2026

Do MLLMs Really See It: Reinforcing Visual Attention in Multimodal LLMs

While chain-of-thought (CoT) reasoning has substantially improved multimodal large language models (MLLMs) on complex reasoning tasks, existing approaches largely rely on long textual reasoning trajectories and provide limited mechanisms for learning stable visual attention policies. Our analysis shows that current MLLMs exhibit weak visual focus: early-stage visual misalignment is rarely corrected during subsequent reasoning, leading to error propagation and failed inferences. We argue that this limitation stems from inadequate credit assignment for visual attention during training. To address this issue, we propose SAYO, a visual reasoning model trained with a reinforcement learning (RL) framework that introduces a region-level visual attention-based reward. This reward explicitly aligns optimization signals with visually grounded reasoning steps, enabling the model to learn more reliable attention behaviors. Extensive experiments across multiple multimodal benchmarks demonstrate that SAYO consistently improves performance on diverse reasoning and perception tasks. Project is open at https://github.com/Cratileo/Sayo.

cs.AI↗
arXiv2026

Observation of measurement-induced localization phenomena in a solid-state spin system

Quantum measurement can profoundly impact the dynamics of many-body systems, for instance, inducing phase transitions from volume-law chaotic phases to area-law localized phases. While chaotic phases emerge naturally in generic many-body systems, the experimental realization of a fully localized phase driven by measurement remains challenging, primarily owing to the stringent requirement of near-perfect isolation of the system from environmental dissipation. Here we experimentally observe measurement-induced localization phenomena in an interacting $^{13}\mathrm{C}$ nuclear spin system in diamond. In contrast to previous circuit-model-based experiments, our well-isolated system evolves under a physical many-body Hamiltonian with measurements intervening at intervals. By implementing either projective measurements on a single spin or global weak measurements on the entire system, we observe substantially prolonged spin autocorrelation times as the measurement frequency increases. Our results, combined with numerical simulations, provide evidence that quantum measurement can drive a many-body system into a localized regime, and lay the groundwork for future experimental investigations into measurement-induced many-body dynamics and related entanglement phenomena.

quant-ph↗
arXiv2026

DocLayout-YOLO: Axis-Attentive Document Layout Detector with Mesh-Guided Synthetic Pretraining

Accurate and efficient document layout analysis is critical for modern document parsing systems. However, although the YOLO framework represents a leading paradigm for the real-time detector, its adaptation to heterogeneous documents remains underexplored. To address this gap, we propose DocLayout-YOLO, a document-specialized real-time detector that integrates diverse synthetic document pretraining and page-axis context modeling. From the data perspective, we introduce MeshLayout-1M, a synthetic pretraining corpus generated by the proposed Mesh-candidate BestFit algorithm, which provides large-scale and diverse document layouts for pretraining and improves adaptation to heterogeneous document scenarios. From the model perspective, we propose Page-Axis Layout Attention (PALA), which captures long-range dependencies along the horizontal and vertical page axes while remaining computationally efficient. Extensive experiments demonstrate that DocLayout-YOLO achieves state-of-the-art performance on four heterogeneous document benchmarks, obtaining 72.4%/75.5%/96.0%/73.1% mAP50:95 on D4LA, M6Doc, HJDataset, and PRIMA-LAD, respectively. Compared with existing real-time detectors, DocLayout-YOLO establishes a superior accuracy-efficiency trade-off in document layout analysis. Code, model, and data are available at https://github.com/opendatalab/DocLayout-YOLO/tree/release-v2.

cs.CV↗
arXiv2026

Exploring the Underwater World Segmentation without Extra Training

Accurate segmentation of marine organisms is vital for biodiversity monitoring and ecological assessment, yet existing datasets and models remain largely limited to terrestrial scenes. To bridge this gap, we introduce \textbf{AquaOV255}, the first large-scale and fine-grained underwater segmentation dataset containing 255 categories and over 20K images, covering diverse categories for open-vocabulary (OV) evaluation. Furthermore, we establish the first underwater OV segmentation benchmark, \textbf{UOVSBench}, by integrating AquaOV255 with five additional underwater datasets to enable comprehensive evaluation. Alongside, we present \textbf{Earth2Ocean}, a training-free OV segmentation framework that transfers terrestrial vision--language models (VLMs) to underwater domains without any additional underwater training. Earth2Ocean consists of two core components: a Geometric-guided Visual Mask Generator (\textbf{GMG}) that refines visual features via self-similarity geometric priors for local structure perception, and a Category-visual Semantic Alignment (\textbf{CSA}) module that enhances text embeddings through multimodal large language model reasoning and scene-aware template construction. Extensive experiments on the UOVSBench benchmark demonstrate that Earth2Ocean achieves significant performance improvement on average while maintaining efficient inference.

cs.CV↗
arXiv2026

Perturbation-Regularized Open-Vocabulary Remote Sensing Segmentation with Unified Multi-Domain Evaluation

Open-vocabulary remote sensing image segmentation (OVRSIS) aims to segment text-specified categories beyond a fixed label space. Its key challenge is to maintain reliable pixel--text correspondence under substantial appearance variation in remote-sensing imagery. Existing methods often rely on specialized vision--language representations or complex adaptation mechanisms, yet still perform deterministic matching between fixed text prototypes and individual visual features. We instead formulate OVRSIS as a \emph{distributional pixel--text alignment} problem and propose \textbf{Pi-Seg}, a lightweight perturbation-based framework. Pi-Seg introduces learnable variations into textual prototypes and dense visual features before alignment. Segmentation supervision encourages perturbations that enlarge the target-to-distractor margin while suppressing harmful feature shifts. Pi-Seg consistently improves performance under the existing \textit{single-source} OVRSISBench protocols. To evaluate OVRSIS beyond source-specific training settings, we further introduce a unified \textit{multi-source} protocol and construct \textbf{GlobalRSOV95K}, containing approximately 95K densely annotated images and 35 semantic categories. Models are trained on this shared dataset and evaluated on an image-disjoint suite of 10 downstream datasets. The complete benchmark covers more than 170K images and 122 categories. Experiments under both single-source and multi-source protocols show that Pi-Seg improves cross-dataset transfer and generalizes effectively to practical geospatial segmentation tasks, while GlobalRSOV95K provides a common foundation for transferable OVRSIS research. \footnote{\url{https://github.com/LiBingyu01/Pi-Seg}}

cs.CV↗
arXiv2026

Viscoelasticity reshapes the frequency response of a rotating magnetic particle

A magnetic particle driven by a rotating magnetic field undergoes a transition from synchronous to asynchronous rotation at a critical driving frequency. The asynchronous dynamics is well understood in Newtonian fluids but remains unclear in viscoelastic media. Here, we develop a theoretical description of the asynchronous rotation of a magnetic particle in a Jeffreys-type viscoelastic fluid. The particle's time-averaged angular velocity exhibits a nontrivial frequency dependence that changes from non-monotonic to monotonic as the polymer relaxation time increases. This behavior is explained by the interplay among magnetic driving, viscoelastic relaxation, and frequency-dependent viscous dissipation. We further derive an asymptotic expression that captures the non-monotonic dependence. These results clarify how solvent and polymer contributions jointly control asynchronous rotation and provide a physical basis for guiding relevant applications in complex fluids.

cond-mat.soft↗
arXiv2026

HERO: Human-profile Enhanced Retrieval Optimization Framework for Long-term Agent Memory

Long-term memory is crucial for personalized responses and long-horizon agent interactions. Existing methods often rely on LLMs to compress or rewrite dialogue histories and use the transformed memories as retrieval evidence. Despite the progress in organizing fragmented contexts, two major drawbacks persist: (1) information loss from compression, which discards fine-grained but later useful details, and (2) semantic drift from rewriting, which erodes the original tone and situated context. In this work, we propose a novel Human-profile Enhanced Retrieval Optimization framework for long-term agent memory (HERO). Specifically, HERO converts the dialogue history into a traceable heterogeneous memory graph that preserves raw dialogue text as evidence for reasoning, thereby mitigating information loss. For retrieval, HERO extracts initial anchors from the current query and incorporates human profiles via an iterative graph traversal; these anchors and profiles provide guidance signals that adaptively activate the most informative regions of the graph. Experiments on two benchmark datasets show that HERO outperforms strong baselines on both factual and personalized reasoning, while providing more faithful access to raw dialogue evidence.

cs.AI↗
arXiv2026

Exploring Efficient Open-Vocabulary Segmentation in the Remote Sensing

Open-Vocabulary Remote Sensing Image Segmentation (OVRSIS), an emerging task that adapts Open-Vocabulary Segmentation (OVS) to the remote sensing (RS) domain, remains underexplored due to the absence of a unified evaluation benchmark and the domain gap between natural and RS images. To bridge these gaps, we first establish a standardized OVRSIS benchmark (\textbf{OVRSISBench}) based on widely-used RS segmentation datasets, enabling consistent evaluation across methods. Using this benchmark, we comprehensively evaluate several representative OVS/OVRSIS models and reveal their limitations when directly applied to remote sensing scenarios. Building on these insights, we propose \textbf{RSKT-Seg}, a novel open-vocabulary segmentation framework tailored for remote sensing. RSKT-Seg integrates three key components: (1) a Multi-Directional Cost Map Aggregation (RS-CMA) module that captures rotation-invariant visual cues by computing vision-language cosine similarities across multiple directions; (2) an Efficient Cost Map Fusion (RS-Fusion) transformer, which jointly models spatial and semantic dependencies with a lightweight dimensionality reduction strategy; and (3) a Remote Sensing Knowledge Transfer (RS-Transfer) module that injects pre-trained knowledge and facilitates domain adaptation via enhanced upsampling. Extensive experiments on the benchmark show that RSKT-Seg consistently outperforms strong OVS baselines by +3.8 mIoU and +5.9 mACC, while achieving 2x faster inference through efficient aggregation. Our code is \href{https://github.com/LiBingyu01/RSKT-Seg}{\textcolor{blue}{here}}.

cs.CV↗
arXiv2026

Semantic-Aware Temporal Adaptation for UAV Anti-UAV Tracking

UAV Anti-UAV tracking is an emerging low-altitude security task for localizing an adversarial UAV using the onboard camera of a moving observer UAV. It differs from conventional UAV tracking and ground-based Anti-UAV tracking because both the camera platform and the target move simultaneously. This dual-dynamic setting induces rapid viewpoint changes, motion blur, scale variation, and visually similar distractors, making reliable appearance matching difficult. Under such rapidly changing conditions, fixed visual representations are often insufficient because target appearance becomes unreliable and feature distributions may deviate from the training domain. The target language description remains stable across frames and can therefore serve as a semantic anchor for temporal state propagation, while online feature-distribution alignment can reduce video-specific test-time shifts. In this paper, we propose \emph{SATATrack}, a Semantic-Aware Temporal Adaptation framework for UAV Anti-UAV tracking. SATATrack introduces Semantic-Aware Context Propagation (SACP), which uses the target description to guide temporal context propagation across backbone stages and preserve target identity under rapid appearance changes. An auxiliary contrastive regularizer is used during training to discourage responses to semantically similar background regions. During inference, Temporal-Aware Distribution Alignment (TADA) aligns feature distributions online without updating model parameters, combining recent-frame estimates with training-time statistics for stability. SATATrack achieves state-of-the-art performance on the UAV-Anti-UAV benchmark while remaining competitive in Anti-UAV and UAV object tracking tasks. The code will be available at https://github.com/XiaozhenQiao/SATATrack.

cs.CV↗
arXiv2026

Molecular chiral discrimination through symmetry-breaking spin dynamics

Molecular chirality plays a crucial role in physics, chemistry, life sciences and pharmacology. Nowadays, the chiral discrimination and control at the single-molecule level is urgently needed to reveal the origin of the chirality-relevant phenomena by recovering the information disturbed by the ensemble averaging. The method of magnetic resonance (MR), as one of powerful tools for structure analysis, is blind to the molecular chirality in the absence of a chiral reagent. Here we propose and experimentally demonstrate a direct MR-based method for determining the chirality at the single-molecule level through constructing the symmetry-breaking dynamics of nearby nuclear spins. In principle, the mirror asymmetry of two enantiomers in real space is manifested by breaking the joint symmetry of the mirror reflection and time reversal in spin space under spin dynamics. Experimentally, two enantiomers are indistinguishable from the dynamics of strongly-coupled but unpolarized nuclear spins, but diverge evidently in the dynamical results that break the field-inversion symmetry after spins are polarized. Our method and results will benefit the study of chirality-induced properties in the fields of chemistry and biology.

quant-ph↗
arXiv2026

MLLM-DataEngine: Closing the Loop of Multimodal Instruction Tuning Data Generation

In this paper, we propose MLLM-DataEngine, a novel closed-loop system that bridges data generation, model training, and evaluation. Within each loop iteration, the MLLM-DataEngine first analyzes the weakness of the model based on the evaluation results, then generates a proper incremental dataset for the next training iteration, and enhances the model capability iteratively. Compared with previous instruction fine-tuning dataset collection methods which are separate from the benchmarking, MLLM-DataEngine shows better targeting and can improve MLLMs's capabilities more effectively. Firstly, we propose an Adaptive Bad-case Sampling module, which can effectively analyze model weakness based on the benchmarking results and adjust the generation of incremental datasets flexibly. Secondly, in order to ensure high-quality data for specific capability types, the most representative in-context examples and abundant information are provided to GPT-4, which helps GPT-4 fully comprehend the model's weakness and further guarantees high-quality generated data. Through extensive experiments, we find MLLM-DataEngine could boost the MLLMs capability in a targeted and automatic manner without human participants. We hope MLLM-DataEngine could be a general solution for the following MLLMs data curation. Code, data, and model are available at https://github.com/opendatalab/MLLM-DataEngine.

cs.MM↗
arXiv2026

Effect of a Cone Shape on the Motion of Active Janus Colloids

The propulsion of active colloids is governed by their symmetry and shape, yet a systematic investigation of how small geometric effects influence the locomotion of anisotropic active colloids is lacking. In this paper, we study the effect of a cone shape by combining high-resolution two-photon polymerization 3D printing with AC electric field experiments to study anisotropic Janus micro swimmers. We fabricate a series of Janus colloids with different shapes, ranging from a sphere to a hemisphere with a cone-protrusion, but containing similar hemispherical gold-coatings. Under an applied alternating current (AC) electric field, all particles exhibit active, ballistic motion. We find a shape and size dependence on the propulsion velocity, with the spherical Janus particle moving fastest, followed by a small cone and a large cone protrusion, while surprisingly a printed sphere with a flat side moves slowest. In addition, a reversal in the direction of motion for all shapes was triggered, a phenomenon governed by a transition from induced-charge electrophoresis (ICEP) at low frequencies to self-dielectrophoresis (sDEP) at high frequencies. We reveal a distinct shape influence, with the large cone-protrusion increasing their velocity in the sDEP regime. Our results provide insight into the link between active particle geometry and their propulsion velocity that are important for understanding biological micro swimmers and designing optimized microrobots.

cond-mat.soft↗
arXiv2026

ChatUMM: Robust Context Tracking for Conversational Interleaved Generation

Unified multimodal models (UMMs) have achieved remarkable progress yet remain constrained by a single-turn interaction paradigm, effectively functioning as solvers for independent requests rather than assistants in continuous dialogue. To bridge this gap, we present ChatUMM. As a conversational unified model, it excels at robust context tracking to sustain interleaved multimodal generation. ChatUMM derives its capabilities from two key innovations: an interleaved multi-turn training strategy that models serialized text-image streams as a continuous conversational flow, and a systematic conversational data synthesis pipeline. This pipeline transforms a diverse set of standard single-turn datasets into fluid dialogues through three progressive stages: constructing basic stateful dialogues, enforcing long-range dependency resolution via ``distractor'' turns with history-dependent query rewriting, and synthesizing naturally interleaved multimodal responses. Extensive evaluations demonstrate that ChatUMM achieves state-of-the-art performance among open-source unified models on visual understanding and instruction-guided editing benchmarks, while maintaining competitive fidelity in text-to-image generation. Notably, ChatUMM exhibits superior robustness in complex multi-turn scenarios, ensuring fluid, context-aware dialogues.

cs.CV↗
arXiv2026

An Open-Source Benchmark and Baseline for Multi-temporal Referring Segmentation

Large Vision-Language Models (LVLMs) have shown strong visual understanding and language-guided grounding abilities, yet their capacity for multi-temporal visual reasoning remains underexplored. To bridge this gap, we introduce \textbf{Multi-temporal Referring Segmentation (MTRS)}, a new task that aims to segment language-described temporal changes from multi-temporal images. MTRS extends conventional referring segmentation and change detection by jointly requiring temporal correspondence reasoning, language grounding, and pixel-level mask prediction. We propose \textbf{CRAFT-Agent}, an automated data construction pipeline with human auditing, and build \textbf{MTRefSeg-21K}, the first MTRS benchmark, containing 21K high-quality multi-temporal image-text-mask triplets across diverse scenes, viewpoints, and domains. Benchmarking a broad set of VLM- and LVLM-based models reveals that direct inference performs poorly, while task-specific fine-tuning remains limited. To address this, we propose \textbf{MTRefSeg-R1}, a change-aware LVLM framework trained with a two-stage strategy. It first learns general temporal-change perception from 20K vision-only bi-temporal samples, and is then fine-tuned on MTRefSeg-21K for fine-grained language-guided temporal localization. MTRefSeg-R1 explicitly models cross-temporal visual differences, aligns language instructions with temporal variations, and predicts referred change masks. Extensive experiments show that MTRefSeg-R1 achieves strong and often superior performance compared with existing LVLM baselines, demonstrating the challenge and potential of MTRS.

cs.CV↗
arXiv2026

MedMamba: Multi-View State Space Models with Adaptive Graph Learning for Medical Time Series Classification

Medical time series are central to healthcare, enabling continuous monitoring and supporting timely clinical decisions. Despite recent progress, existing methods struggle to jointly model local-global dynamics and handle nonstationarities like baseline drift, while often failing to capture latent channel interactions. To address these challenges, we propose MedMamba, an end-to-end architecture that integrates state space models with domain-specific inductive biases. Specifically, MedMamba first employs multi-scale convolutional embeddings to capture discriminative local morphology. Second, to mitigate nonstationarity, we introduce a tri-branch differential state space encoder that processes raw, temporal-difference, and frequency-domain views, fusing them to emphasize informative patterns while suppressing drift. Furthermore, to uncover latent channel correlations, we design a spatial graph Mamba module that learns a directed dependency structure regularized toward sparsity and acyclicity, which obviates the need for predefined graphs. Extensive experiments on five real-world datasets demonstrate that MedMamba achieves state-of-the-art performance while maintaining linear computational complexity, and ablation studies validate each component's contribution.Code is available at https://github.com/zhangda1018/MedMamba.

cs.LG↗
arXiv2026

A Visual Reinforcement Learning-Based Separate Primitive Policy for Peg-in-Hole Tasks

For peg-in-hole tasks, humans rely on binocular visual perception to locate the peg above the hole surface and then proceed with insertion. This paper draws insights from this behavior to enable agents to learn efficient assembly strategies through visual reinforcement learning. Hence, we propose a Separate Primitive Policy (S2P) to learn how to derive location and insertion actions simultaneously. S2P is compatible with model-free reinforcement learning algorithms. Ten insertion tasks featuring different polygons are developed as benchmarks for evaluations. Simulation experiments show that S2P can boost the sample efficiency and success rate even with force constraints. Real-world experiments are also performed to verify the feasibility of S2P. Ablations are finally given to discuss the generalizability of S2P and some factors that affect its performance.

cs.RO↗