arXiv subjects

Zihao Zhou

Publications and source records attributed to Zihao Zhou.

At least 19 recordsLinked to original sources

arXiv2026

Motus2: A Self-Evolving General World Model for Dexterous Manipulation

General embodied agents should perceive, predict, act, evaluate, and improve within a unified system. World models have shown great promise in building such agents, yet existing models typically append an action output head to a world simulator, without coupling them into a closed decision-and-learning loop for policy improvement. We present Motus2, a self-evolving general world model for dexterous manipulation. Motus2 advances world modeling through model scaling and data scaling. For model scaling, a single model with shared weights exposes three control interfaces: a policy (world-action model), a simulator (action-conditioned world model), and an evaluator (value model). The policy proposes candidate action chunks, the simulator predicts their visual consequences, and the evaluator assesses the predicted outcomes. Their coupling forms a closed decision-and-learning loop for policy improvement. This formulation uses curated expert demonstrations for action learning, while failed and suboptimal interactions provide valuable evidence for dynamics modeling and value learning. For data scaling, Motus2 progresses from large-scale monocular egocentric data to synchronized stereo egocentric data, followed by robot-domain adaptation with robot trajectories and supplementary human-robot alignment data. Motus2 further studies global-autoregressive and hybrid-memory extensions of its sliding-window context, adds tactile feedback for contact-aware control, and is instantiated on a fully biomimetic platform with stereo vision, dual arms, dual dexterous hands, and tactile sensing. Together, egocentric data scaling and closed-loop general world model scaling provide a general path toward self-evolving dexterous manipulation.

cs.RO↗

arXiv2026

DiaRelay: Relaying Dialogue Context with a Constant-Size Memory for Emotion Recognition in Conversation

Emotion Recognition in Conversation (ERC) requires models to identify subtle emotional cues that are often distributed across distant dialogue turns. Existing methods typically incorporate dialogue history through a fixed context window. However, short windows discard potentially useful long-range evidence, while enlarging the window repeatedly re-encodes overlapping utterances, increases computational and memory costs, and may introduce irrelevant context. Moreover, commonly used parameter-efficient adaptation methods, such as LoRA, mainly introduce fixed low-rank transformations in the feature space and do not explicitly maintain a dialogue-level state or condition their transformations on the evolving conversational context. To address these limitations, we propose a lightweight adapter, DiaRelay, to enable LLMs to explicitly maintain a dialogue-level memory for accurate ERC. Based on LoRA, DiaRelay introduces two extra tightly collaborative components, Selective Relay Memory Transition and Dual-axis Relay Memory Read. Selective Relay Memory Transition progressively aggregates useful historical evidence into a bounded relay memory and propagates it across successive utterance predictions. This allows earlier emotional cues to influence later predictions after they leave the local context window, without re-encoding the complete dialogue history or expanding the backbone context length. Dual-axis Relay Memory Read uses the propagated memory to dynamically modulate low-rank feature transformations, enabling context-dependent representation adaptation without test-time gradient updates. Extensive experiments show that DiaRelay can achieve SOTA weighted F1 and accuracy on MELD while obtaining competitive results on IEMOCAP with only an extra 7.1M trainable parameters, indicating the effectiveness and generalizability of our DiaRelay in enhancing LLM-based emotional understanding.

cs.CL↗

arXiv2026

Fidelity Where it Matters: Site-Specific Nonuniform Refinement for Wireless Digital Twins

Wireless digital twins (WDTs) enable site-specific learning, management, and evaluation in wireless networks. However, constructing and maintaining a high-fidelity WDT over large-scale complex environments can be prohibitively expensive, especially in terms of data acquisition, geometric reconstruction, storage, and ray tracing. To address this issue, a task-oriented nonuniform refinement framework for WDTs is proposed, where limited resources are selectively allocated to the WDT components that matter most to wireless fidelity. Specifically, a unified refinement framework is first developed, which maximizes task-level fidelity under resource constraints through fine-grained component-wise fidelity allocation. This framework is then instantiated for building-level geometry refinement in urban WDTs. It is found that different buildings exhibit highly heterogeneous impacts on wireless fidelity. Motivated by this observation, an ellipsoid-guided selective refinement algorithm (EGSR) is proposed. By jointly considering the relevance of each building to both line-of-sight (LoS) and non-line-of-sight (NLoS) propagation paths, its refinement priority can be estimated using only a low-fidelity WDT. Simulations across multiple urban scenarios show that EGSR can substantially improve radio-map fidelity and preserve beamforming effectiveness by refining only a small subset of buildings. These results demonstrate the potential of task-oriented fidelity allocation as a scalable principle for constructing efficient and performance-aware WDTs, thereby facilitating reliable site-specific learning and optimization.

eess.SP↗

arXiv2026

GLM-5V-Turbo: Toward a Native Foundation Model for Multimodal Agents

We present GLM-5V-Turbo, a step toward native foundation models for multimodal agents. As foundation models are increasingly deployed in real environments, agentic capability depends not only on language reasoning, but also on the ability to perceive, interpret, and act over heterogeneous contexts such as images, videos, webpages, documents, GUIs. GLM-5V-Turbo is built around this objective: multimodal perception is integrated as a core component of reasoning, planning, tool use, and execution, rather than as an auxiliary interface to a language model. This report summarizes the main improvements behind GLM-5V-Turbo across model design, multimodal training, reinforcement learning, toolchain expansion, and integration with agent frameworks. These developments lead to strong performance in multimodal coding, visual tool use, and framework-based agentic tasks, while preserving competitive text-only coding capability. More importantly, our development process offers practical insights for building multimodal agents, highlighting the central role of multimodal perception, hierarchical optimization, and reliable end-to-end verification.

cs.CV↗

arXiv2026

Hierarchical Alignment: Enforcing Hierarchical Instruction-Following in LLMs through Logical Consistency

Large language models increasingly operate under multiple instructions from heterogeneous sources with different authority levels, including system policies, user requests, tool outputs, and retrieved context. While prior work on instruction hierarchy highlights the importance of respecting instruction priorities, it mainly focuses on adversarial attacks and overlooks the benign but common instruction conflicts that arise in real-world applications. In such settings, models must not only avoid security violations but also preserve task utility and behavioral consistency when instructions partially or implicitly conflict. We propose Neuro-Symbolic Hierarchical Alignment (NSHA) for hierarchical instruction-following by explicitly modeling and enforcing instruction priorities. At inference time, we introduce solver-guided reasoning that formulates instruction resolution as a constraint satisfaction problem, enabling the model to derive a maximally consistent set of applicable instructions under hierarchical constraints. At training time, NSHA distills solver-based decisions into model parameters using automatically constructed supervision. We evaluate our approach on rule following, task execution, tool use, and safety, covering both single-turn and multi-turn interactions, and show that NSHA significantly improves performance under such conflicts while maintaining competitive utility in reference settings.

cs.CL↗

arXiv2026

EruDiff: Refactoring Knowledge in Diffusion Models for Advanced Text-to-Image Synthesis

Text-to-image diffusion models have achieved remarkable fidelity in synthesizing images from explicit text prompts, yet exhibit a critical deficiency in processing implicit prompts that require deep-level world knowledge, ranging from natural sciences to cultural commonsense, resulting in counter-factual synthesis. This paper traces the root of this limitation to a fundamental dislocation of the underlying knowledge structures, manifesting as a chaotic organization of implicit prompts compared to their explicit counterparts. In this paper, we propose EruDiff, which aims to refactor the knowledge within diffusion models. Specifically, we develop the Diffusion Knowledge Distribution Matching (DK-DM) to register the knowledge distribution of intractable implicit prompts with that of well-defined explicit anchors. Furthermore, to rectify the inherent biases in explicit prompt rendering, we employ the Negative-Only Reinforcement Learning (NO-RL) strategy for fine-grained correction. Rigorous empirical evaluations demonstrate that our method significantly enhances the performance of leading diffusion models, including FLUX and Qwen-Image, across both the scientific knowledge benchmark (i.e., Science-T2I) and the world knowledge benchmark (i.e., WISE), underscoring the effectiveness and generalizability. Our code is available at https://github.com/xiefan-guo/erudiff.

cs.CV↗

arXiv2026

Fast Beam-Brainstorm: Few-Step Generative Site-Specific Beamforming with Flexible Probing

A novel generative site-specific beamforming (GenSSBF) approach, termed fast beam-brainstorm (F-BBS), is proposed to address the practical bottlenecks of slow beam generation and fixed channel probing lengths in existing GenSSBF. To accelerate beam generation, F-BBS utilizes a two-stage distillation strategy that learns an average velocity field, instead of an instantaneous one, to guide the beam generative process. This strategy enables larger generation steps, realizing few-step or even one-step beam generation. Furthermore, to accommodate flexible channel probing lengths, a stochastic masking mechanism and a beam index-aware masked condition encoder are proposed, enabling a single trained model to operate with variable-length channel probing observations without retraining. Therefore, FBBS achieves the fast generation of high-fidelity communication beams from coarse and variable-length channel probing feedback, i.e., reference signal received power (RSRP), from user equipments. Simulation results on accurate ray-tracing datasets show that 1) F-BBS achieves comparable performance while reducing the beam generation cost by over 90% compared with diffusion-based GenSSBF solutions, 2) F-BBS realizes robust performance across variable channel probing length, and 3) FBBS offers a desirable trade-off between beamforming gain and beam probing overhead.

eess.SP↗

arXiv2026

SimulCost: A Cost-Aware Benchmark and Toolkit for Automating Physics Simulations with LLMs

Evaluating LLM agents for scientific tasks has focused on token costs while ignoring tool-use costs like simulation time and experimental resources. As a result, metrics like pass@k become impractical under realistic budget constraints. To address this gap, we introduce SimulCost, the first benchmark targeting cost-sensitive parameter tuning in physics simulations. SimulCost compares LLM tuning cost-sensitive parameters against traditional scanning approach in both accuracy and computational cost, spanning 2,643 single-round (initial guess) and 2,304 multi-round (adjustment by trial-and-error) tasks across 11 simulators from fluid dynamics, solid mechanics, and plasma physics, whose costs are analytically defined and platform-independent. A twelfth simulator, a production plasma code measurable only by wall clock, is reported separately. Frontier LLMs achieve 45-62% success rates in single-round mode, dropping to 34-50% under high accuracy requirements, rendering their initial guesses unreliable especially for high accuracy tasks. Multi-round mode improves rates to 66-81%, but LLMs are 1.5-2.7x slower than traditional scanning, making them uneconomical choices. We also investigate parameter group correlations for knowledge transfer potential, and the impact of in-context examples and reasoning effort, providing practical implications for deployment and fine-tuning. We open-source SimulCost as a static benchmark and extensible toolkit to facilitate research on improving cost-aware agentic designs for physics simulations, and for expanding new simulation environments. Code and data are available at https://github.com/Rose-STL-Lab/SimulCost-Bench

physics.comp-ph↗

arXiv2026

Beyond Benchmarks of IUGC: Rethinking Requirements of Deep Learning Methods for Intrapartum Ultrasound Biometry from Fetal Ultrasound Videos

A substantial proportion (45\%) of maternal deaths, neonatal deaths, and stillbirths occur during the intrapartum phase, with a particularly high burden in low- and middle-income countries. Intrapartum biometry plays a critical role in monitoring labor progression; however, the routine use of ultrasound in resource-limited settings is hindered by a shortage of trained sonographers. To address this challenge, the Intrapartum Ultrasound Grand Challenge (IUGC), co-hosted with MICCAI 2024, was launched. The IUGC introduces a clinically oriented multi-task automatic measurement framework that integrates standard plane classification, fetal head-pubic symphysis segmentation, and biometry, enabling algorithms to exploit complementary task information for more accurate estimation. Furthermore, the challenge releases the largest multi-center intrapartum ultrasound video dataset to date, comprising 774 videos (68,106 frames) collected from three hospitals, providing a robust foundation for model training and evaluation. In this study, we present a comprehensive overview of the challenge design, review the submissions from eight participating teams, and analyze their methods from five perspectives: preprocessing, data augmentation, learning strategy, model architecture, and post-processing. In addition, we perform a systematic analysis of the benchmark results to identify key bottlenecks, explore potential solutions, and highlight open challenges for future research. Although encouraging performance has been achieved, our findings indicate that the field remains at an early stage, and further in-depth investigation is required before large-scale clinical deployment. All benchmark solutions and the complete dataset have been publicly released to facilitate reproducible research and promote continued advances in automatic intrapartum ultrasound biometry.

cs.CV↗

arXiv2026

FUGC: Benchmarking Semi-Supervised Learning Methods for Cervical Segmentation

Accurate segmentation of cervical structures in transvaginal ultrasound (TVS) is critical for assessing the risk of spontaneous preterm birth (PTB), yet the scarcity of labeled data limits the performance of supervised learning approaches. This paper introduces the Fetal Ultrasound Grand Challenge (FUGC), the first benchmark for semi-supervised learning in cervical segmentation, hosted at ISBI 2025. FUGC provides a dataset of 890 TVS images, including 500 training images, 90 validation images, and 300 test images. Methods were evaluated using the Dice Similarity Coefficient (DSC), Hausdorff Distance (HD), and runtime (RT), with a weighted combination of 0.4/0.4/0.2. The challenge attracted 10 teams with 82 participants submitting innovative solutions. The best-performing methods for each individual metric achieved 90.26\% mDSC, 38.88 mHD, and 32.85 ms RT, respectively. FUGC establishes a standardized benchmark for cervical segmentation, demonstrates the efficacy of semi-supervised methods with limited labeled data, and provides a foundation for AI-assisted clinical PTB risk assessment.

eess.IV↗

arXiv2026

Numina-Lean-Agent: An Open and General Agentic Reasoning System for Formal Mathematics

Agentic systems have recently become the dominant paradigm for formal theorem proving, achieving strong performance by coordinating multiple models and tools. However, existing approaches often rely on task-specific pipelines and trained formal provers, limiting their flexibility and reproducibility. In this paper, we propose the paradigm that directly uses a general coding agent as a formal math reasoner. This paradigm is motivated by (1) A general coding agent provides a natural interface for diverse reasoning tasks beyond proving, (2) Performance can be improved by simply replacing the underlying base model, without training, and (3) MCP enables flexible extension and autonomous calling of specialized tools, avoiding complex design. Based on this paradigm, we introduce Numina-Lean-Agent, which combines Claude Code with Numina-Lean-MCP to enable autonomous interaction with Lean, retrieval of relevant theorems, informal proving and auxiliary reasoning tools. Using Claude Opus 4.5 as the base model, Numina-Lean-Agent solves all problems in Putnam 2025 (12 / 12), matching the best closed-source system. Beyond benchmark evaluation, we further demonstrate its generality by interacting with mathematicians to successfully formalize the Brascamp-Lieb theorem. We release Numina-Lean-Agent and all solutions at https://github.com/project-numina/numina-lean-agent.

cs.AI↗

arXiv2026

Beam-Brainstorm: A Generative Site-Specific Beamforming Approach

Accurately understanding the propagation environment is a fundamental challenge in site-specific beamforming (SSBF). This paper proposes a novel generative SSBF (GenSSBF) solution, which represents a paradigm shift from conventional unstructured prediction to joint-structure modeling. First, considering the fundamental differences between beam generation and conventional image synthesis, a unified GenSSBF framework is proposed, which includes a site profile, a wireless prompting module, and a generator. Second, a beam-brainstorm (BBS) solution is proposed as an instantiation of this GenSSBF framework. Specifically, the site profile is configured by transforming channel data from spatial domain to a reversible latent space via discrete Fourier transform (DFT). To facilitate practical deployment, the wireless prompt is constructed from the reference signal received power (RSRP) measured using a small number of DFT-beams. Finally, the generator is developed using a customized conditional diffusion model. Rather than relying on a meticulously designed global codebook, BBS directly generates diverse and high-fidelity user-specific beams guided by the wireless prompts. Simulation results on accurate ray-tracing datasets demonstrate that BBS can achieve near-optimal beamforming gain while drastically reducing the beam sweeping overhead, even in low signal-to-noise ratio (SNR) environments.

eess.SP↗

arXiv2026

Generative Site-Specific Beamforming for Next-Generation Spatial Intelligence

This article proposes generative site-specific beamforming (GenSSBF) for next-generation spatial intelligence in wireless networks. Site-specific beamforming (SSBF) has emerged as a promising paradigm to mitigate the channel acquisition bottleneck in multiantenna systems by exploiting environmental priors. However, classical SSBF based on discriminative deep learning struggles: 1) to properly represent the inherent multimodality of wireless propagation and 2) to effectively capture the structural features of beamformers. In contrast, by leveraging conditional generative models, GenSSBF addresses these issues via learning a conditional distribution over feasible beamformers. By doing so, the synthesis of diverse and high-fidelity beam candidates from coarse channel sensing measurements can be guaranteed. This article presents the fundamentals, system designs, and implementation methods of GenSSBF. Case studies in both indoor and outdoor scenarios show that GenSSBF attains near-optimal beamforming gain with ultra-low channel acquisition overhead. Finally, several open research problems are highlighted.

cs.IT↗

arXiv2025

FairGFL: Privacy-Preserving Fairness-Aware Federated Learning with Overlapping Subgraphs

Graph federated learning enables the collaborative extraction of high-order information from distributed subgraphs while preserving the privacy of raw data. However, graph data often exhibits overlap among different clients. Previous research has demonstrated certain benefits of overlapping data in mitigating data heterogeneity. However, the negative effects have not been explored, particularly in cases where the overlaps are imbalanced across clients. In this paper, we uncover the unfairness issue arising from imbalanced overlapping subgraphs through both empirical observations and theoretical reasoning. To address this issue, we propose FairGFL (FAIRness-aware subGraph Federated Learning), a novel algorithm that enhances cross-client fairness while maintaining model utility in a privacy-preserving manner. Specifically, FairGFL incorporates an interpretable weighted aggregation approach to enhance fairness across clients, leveraging privacy-preserving estimation of their overlapping ratios. Furthermore, FairGFL improves the tradeoff between model utility and fairness by integrating a carefully crafted regularizer into the federated composite loss function. Through extensive experiments on four benchmark graph datasets, we demonstrate that FairGFL outperforms four representative baseline algorithms in terms of both model utility and fairness.

cs.LG↗

arXiv2025

Neural reconstruction of 3D ocean wave hydrodynamics from camera sensing

Precise three-dimensional (3D) reconstruction of wave free surfaces and associated velocity fields is essential for developing a comprehensive understanding of ocean physics. To address the high computational cost of dense visual reconstruction in long-term ocean wave observation tasks and the challenges introduced by persistent visual occlusions, we propose an wave free surface visual reconstruction neural network, which is designed as an attention-augmented pyramid architecture tailored to the multi-scale and temporally continuous characteristics of wave motions. Using physics-based constraints, we perform time-resolved reconstruction of nonlinear 3D velocity fields from the evolving free-surface boundary. Experiments under real-sea conditions demonstrate millimetre-level wave elevation prediction in the central region, dominant-frequency errors below 0.01 Hz, precise estimation of high-frequency spectral power laws, and high-fidelity 3D reconstruction of nonlinear velocity fields, while enabling dense reconstruction of two million points in only 1.35 s. Built on a stereo-vision dataset, the model outperforms conventional visual reconstruction approaches and maintains strong generalization in occluded conditions, owing to its global multi-scale attention and its learned encoding of wave propagation dynamics.

cs.CV↗

arXiv2025

Spectral and Energy Efficiency Tradeoff for Pinching-Antenna Systems

The joint transmit and pinching beamforming design for spectral efficiency (SE) and energy efficiency (EE) tradeoff in pinching-antenna systems (PASS) is proposed, under practical channel and energy consumption models. In the single-user scenario, it is proved that the optimal pinching antenna (PA) positions are independent of the transmit beamforming. Based on this insight, a two-stage joint beamforming design is proposed. Specifically, in the first stage, a general PA placement framework is proposed for multi-waveguide systems. In the second stage, the closed-form solution for the optimal transmit beamformer is derived given the optimized PA positions. In the multi-user scenario, an alternating optimization (AO)-based joint beamforming design is proposed to balance the SE-EE performance while taking the quality-of-service (QoS) requirements into account. It is proved that the proposed AO-based algorithm is guaranteed to converge when no constraints are violated in PA placement subproblem. Numerical results demonstrate that: 1) the proposed algorithms effectively improve joint SE-EE performance; 2) PASS exhibits strong robustness against variations in the service area along the waveguide direction.

eess.SP↗

arXiv2025

Can MLLMs Absorb Math Reasoning Abilities from LLMs as Free Lunch?

Math reasoning has been one crucial ability of large language models (LLMs), where significant advancements have been achieved in recent years. However, most efforts focus on LLMs by curating high-quality annotation data and intricate training (or inference) paradigms, while the math reasoning performance of multi-modal LLMs (MLLMs) remains lagging behind. Since the MLLM typically consists of an LLM and a vision block, we wonder: Can MLLMs directly absorb math reasoning abilities from off-the-shelf math LLMs without tuning? Recent model-merging approaches may offer insights into this question. However, they overlook the alignment between the MLLM and LLM, where we find that there is a large gap between their parameter spaces, resulting in lower performance. Our empirical evidence reveals two key factors behind this issue: the identification of crucial reasoning-associated layers in the model and the mitigation of the gaps in parameter space. Based on the empirical insights, we propose IP-Merging that first identifies the reasoning-associated parameters in both MLLM and Math LLM, then projects them into the subspace of MLLM, aiming to maintain the alignment, and finally merges parameters in this subspace. IP-Merging is a tuning-free approach since parameters are directly adjusted. Extensive experiments demonstrate that our IP-Merging method can enhance the math reasoning ability of MLLMs directly from Math LLMs without compromising their other capabilities.

cs.AI↗

arXiv2025

CaTS-Bench: Can Language Models Describe Time Series?

Time series captioning, the task of describing time series in natural language, requires numeric and temporal reasoning, trend interpretation, and contextual understanding. Existing benchmarks, however, often rely on fully synthetic or generic captions, and typically neglect metadata and visual representations. We introduce CaTS-Bench, a comprehensive benchmark for Context-aware Time Series reasoning across 11 diverse domains, centered on a gold-standard evaluation set of 1746 human-rewritten captions that measure how effectively models translate numeric trends into immediately interpretable narratives. To address the scarcity of human-annotated data, we also propose a scalable pipeline for generating high-fidelity synthetic captions, the quality of which we validate. We evaluate leading Vision-Language Models on our benchmark, revealing that even proprietary models struggle to capture numeric nuances in temporal descriptions, while finetuning open-source models on synthetic data yields substantial performance gains. Finally, we release a diagnostic suite of 910 multiple-choice questions and use tailored numeric metrics to gauge time-series-specific reasoning capabilities, establishing CaTS-Bench as a reliable foundation for grounded, multimodal text generation in numeric domains.

cs.LG↗