arXiv subjects

Kai Peng

Publications and source records attributed to Kai Peng.

At least 19 recordsLinked to original sources

arXiv2026

SAMOT: State-Aware Step Modulation and Optimal Transport Matching for Audio-Visual Instance Segmentation

Audio-Visual Instance Segmentation (AVIS) aims to simultaneously classify, segment, and track sounding objects within video sequences. Unlike Audio-Visual Semantic Segmentation (AVS), AVIS involves instance-level modeling across longer video sequences, introducing two key challenges: (1) complex modality-state changes disrupt long-range modeling, and (2) substantial structural and distributional discrepancies between modalities hinder precise instance-level association. Existing methods rely on fixed-step Transformers and recursive Mamba models, lacking adaptability to modality-state changes. In addition, methods performing implicit matching ignore the inherent distributional inconsistencies. To address these issues, we propose a framework with Adaptive Dynamic Step Modulation (ADSM) and Optimal Transport-based Matching Modulation (OT-MM). ADSM adaptively modulates Mamba step sizes using temporal variation, cross-modal discrepancy, and historical context, balancing rapid response to modality-state changes with stable long-range modeling. OT-MM explicitly formulates instance-level cross-modal matching as an entropy-regularized optimal transport problem solved via log-domain Sinkhorn iterations, and further enforces distribution-level coherence with an MMD regularizer. Extensive experiments demonstrate state-of-the-art performance on the AVIS benchmark (+3.76 FSLA, +2.75 HOTA, +2.58 mAP), verified through comprehensive qualitative visualizations. The code and model are available at https://github.com/happylife-pk/SAMOT.

cs.MM↗

arXiv2026

Hear to See: Discerning Stateful Listening for Audio-Visual Instance Segmentation

Audio-visual instance segmentation (AVIS) requires accurately identifying and tracking individual sounding objects with pixel-level masks. Existing methods struggle to match overlapping acoustic events with visual instances and handle asynchronous audio-visual dynamics. Therefore, two critical questions arise: how can a model establish precise correspondence between overlapping sound sources and visual instances, and how can a model maintain robust tracking when audio and visual signals are temporally misaligned?This paper proposes Hear to See (H2S), addressing these challenges through two mechanisms. The Acoustic-Semantic Projector (ASP) disentangles mixed audio and establishes hierarchical correspondence from semantic to spatial domains. The Asynchronous Dynamics Modulator (ADM) adaptively adjusts state transitions via audio-modulated Mamba, prioritizing current information during dynamic variations and maintaining continuity in stable periods.Experiments on AVISeg show H2S achieves SOTA performance, attaining 48.54 mAP with a COCO pretrained ResNet50 and surpassing the previous by 7.8\%. The code will be open-sourced once the paper is accepted. The source code will be publicly available at https://github.com/leiyeliu/H2S.

cs.MM↗

arXiv2026

AC-VLA: Robust Out-of-Distribution Action Execution via Compositional Learning

Vision-Language-Action (VLA) models excel at end-to-end robotic manipulation but struggle with out-of-distribution (OOD) generalization when familiar sub-tasks are recombined in unseen configurations. We identify two mutually reinforcing failure modes: \emph{trajectory overfitting}, where models overfit to holistic trajectory patterns rather than compositional sub-skill semantics; and \emph{perceptual shortcut}, where action tokens over-rely on wrist-view textures at the expense of global spatial grounding. To address both, we introduce \textbf{AC-VLA}, a plug-and-play Action Compositional learning framework comprising two architecture-agnostic components: \textbf{(i)} a compositional learning module that uses an LLM-driven instruction decomposer and a proprioceptive trajectory aligner to generate dense sub-task supervision, followed by mixed training on complete demonstrations and decomposed data to endow the model with compositional generalization; and \textbf{(ii)} a state-conditioned asymmetric masking strategy that suppresses wrist-view inputs during closed-gripper phases, enforcing global semantic grounding. All components are architectural modification-free and directly integrable into any VLA backbone. Instantiated on $\pi_{0.5}$ and evaluated on LIBERO and LIBERO-OOD benchmarks, AC-VLA achieves a ~28% absolute improvement on compositional OOD tasks while maintaining near-perfect in-distribution performance.

cs.RO↗

arXiv2026

Unleashing More Actions via Action Compositional Training for VLA Models

Vision-Language-Action models excel at robotic manipulation, driven by the scale and diversity of demonstration data. However, standard training paradigms often cause VLA models to severely overfit to specific behavioral patterns, rendering them unable to generalize to out-of-distribution scenarios even when those scenarios merely require novel combinations of identical sub-skills. While expanding datasets can mitigate this overfitting, acquiring high-quality robot data remains notoriously labor-intensive and cost-prohibitive. To resolve this impasse without expensive human teleoperation and to truly unleash more actions,i.e., enable VLA models to compose known sub-skills into a much broader set of executable behaviors beyond the original demonstrations-we propose ACT-VLA (Action Compositional Training for VLA Models), an offline data augmentation framework that leverages the model's latent task representations to synthesize novel, physically valid demonstrations directly from existing tasks for policy training. By eliminating additional manual data collection, our method automatically expands the training distribution and mitigates overfitting. We evaluate our approach on challenging manipulation tasks in simulation. Experiments demonstrate that while baseline VLA models generalize poorly due to original distribution overfitting, policies trained with our synthesized data achieve substantially higher success rates, validating that leveraging existing tasks for automated demonstration synthesis provides an effective, scalable, and data-efficient route to broadening VLA generalization.

cs.RO↗

arXiv2026

Selective Noise Suppression and Discriminative Mutual Interaction for Robust Audio-Visual Segmentation

The ability to capture and segment sounding objects in dynamic visual scenes is crucial for the development of Audio-Visual Segmentation (AVS) tasks. While significant progress has been made in this area, the interaction between audio and visual modalities still requires further exploration. In this work, we aim to answer the following questions: How can a model effectively suppress audio noise while enhancing relevant audio information? How can we achieve discriminative interaction between the audio and visual modalities? To this end, we propose SDAVS, equipped with the Selective Noise-Resilient Processor (SNRP) module and the Discriminative Audio-Visual Mutual Fusion (DAMF) strategy. The proposed SNRP mitigates audio noise interference by selectively emphasizing relevant auditory cues, while DAMF ensures more consistent audio-visual representations. Experimental results demonstrate that our proposed method achieves state-of-the-art performance on benchmark AVS datasets, especially in multi-source and complex scenes. \textit{The code and model are available at https://github.com/happylife-pk/SDAVS}.

cs.CV↗

arXiv2025

Experimental realization of a full-band wave antireflection based on temporal taper metamaterials

As time can be introduced as an additional degree of freedom, temporal metamaterials nowadays open up new avenues for wave control and manipulation. Among these advancements, temporal metamaterial-based antireflection coatings have recently emerged as an innovative method that inherently avoids additional spatial insertions. However, prior temporal antireflection models with finite inserted temporal transition sections that rely on the destructive interference mechanism exhibit residual periodic strong reflections at high frequencies, fundamentally limiting the achievable bandwidth. In this work, the concept of "temporal taper", the temporal counterpart of a conventional spatial taper with a nearly full-band antireflection feature and good compatibility with gradual time-varying components, has been experimentally realized. A 1D temporal metamaterial base on voltage-controlled varactors has been designed experimentally validated. The temporal taper based broadband antireflection exempts the system from spatial matching insertions, and enables agile impedance matching for various terminal loads, positioning it as a promising approach in future photonic systems.

physics.optics↗

arXiv2025

Frequency-Domain Decomposition and Recomposition for Robust Audio-Visual Segmentation

Audio-visual segmentation (AVS) plays a critical role in multimodal machine learning by effectively integrating audio and visual cues to precisely segment objects or regions within visual scenes. Recent AVS methods have demonstrated significant improvements. However, they overlook the inherent frequency-domain contradictions between audio and visual modalities--the pervasively interfering noise in audio high-frequency signals vs. the structurally rich details in visual high-frequency signals. Ignoring these differences can result in suboptimal performance. In this paper, we rethink the AVS task from a deeper perspective by reformulating AVS task as a frequency-domain decomposition and recomposition problem. To this end, we introduce a novel Frequency-Aware Audio-Visual Segmentation (FAVS) framework consisting of two key modules: Frequency-Domain Enhanced Decomposer (FDED) module and Synergistic Cross-Modal Consistency (SCMC) module. FDED module employs a residual-based iterative frequency decomposition to discriminate modality-specific semantics and structural features, and SCMC module leverages a mixture-of-experts architecture to reinforce semantic consistency and modality-specific feature preservation through dynamic expert routing. Extensive experiments demonstrate that our FAVS framework achieves state-of-the-art performance on three benchmark datasets, and abundant qualitative visualizations further verify the effectiveness of the proposed FDED and SCMC modules. The code will be released as open source upon acceptance of the paper.

cs.CV↗

arXiv2025

Terahertz-Driven Nano-tip Field-Emission Electron Gun and Cascaded Acceleration

This paper reports two versions of terahertz (THz)-driven nanotip field-emission electron guns: single-layer reflective guns (SLRGs) and double-layer reflective guns (DLRGs). Both guns use nanotip emitters and accelerate electrons through the electric field of the THz wave. SLRGs employ a reflective structure to superimpose the initial and subsequent half-cycles of the THz electric field, enhancing the field amplitude and acceleration efficiency. Experiments have demonstrated that SLRGs achieve higher acceleration efficiency than single-layer nonreflective guns (SLNRGs) for identical THz input energies. This constitutes direct experimental verification of the efficacy of the reflective structure. Theoretically, SLRGs operating in single-feed mode can match the acceleration efficiency of dual-feed SLNRGs while reducing operational complexity. DLRGs demonstrate THz-driven cascaded electron acceleration through precise scanning of the delay between two incident THz beams. This represents a direct experimental demonstration of cascaded acceleration in THz-driven electron sources. The experimental results of DLRGs align closely with the results of electron dynamics predicted by simulations. This establishes the foundation for developing multilayer high-acceleration-efficiency THz-driven high-energy electron guns. The ability to manipulate the THz for each layer individually holds promising potential for improving the beam quality of THz electron guns.

physics.acc-ph↗

arXiv2025

Observation and Interpretation of Field Emission Saturation Induced by an Ultra-fast Intense Terahertz Field

Field emission under ultra-fast intense terahertz fields provides a promising approach for generating electron bunches with ultrashort pulse duration and high charge densities. It is generally believed that the field emission current described by traditional field emission theory increases dramatically with the applied electric field. However, we conducted extensive field emission experiments using quasi-single-cycle strong-field terahertz radiation at various energy levels and different temperatures and observed an intriguing phenomenon where the emitted charge reached saturation. A novel model is proposed to interpret this phenomenon, which considers the contribution of surface valence electrons and the dynamic replenishment of free electrons from the bulk to the surface. The experimentally observed convex relationship between the emitted charge and terahertz energy is consistent with the model prediction, unlike the concave relationship derived from the traditional field emission formula. In addition, another observed counter-intuitive phenomenon, the inverse correlation between the cathode temperature and saturated emission charge, is also well interpreted by the model. This work offers comprehensive insights into field emission dynamics under ultra-fast intense fields, paving the way for generating electron bunches with unprecedented temporal resolution.

physics.acc-ph↗

arXiv2024

A Lightweight and Scalable Design of Segment Routing in Broadband LEO Constellations Using Landmark-Based Skeleton Graphs

Emerging Low Earth Orbit (LEO) broadband constellations hold significant potential to provide advanced Internet services due to inherent geometric features of the grid topology. However, high dynamics, unstable topology changes, and frequent route updates bring significant challenge to fast and adaptive routing policies. In addition, since computing, bandwidth, and storage resources in each LEO satellite is strictly limited, traffic demands are typically unbalanced, further enlarging the challenge to scalable routing policies with load balancing. Nevertheless, most existing research failed to address the above difficulties. Therefore, this paper proposes a lightweight and scalable protocol of segment routing through landmark-based skeleton graphs. To improve the overall performance, we design an efficient multipath segment routing algorithm. First, the algorithm partitions the network into multiple regions to construct skeleton paths, which can effectively guide packet forwarding and reduce the operating costs. In each region, multipath probabilistic routing is used to achieve uniform traffic distribution, avoiding hotspot congestion. Furthermore, the flexible hierarchical partitioning and localized segmented routing is employed for fine-grained traffic control and QoS guarantee combined with adaptive local single-path routing. Finally, experimental results validate our method's superior performance in terms of response time and network utility.

cs.NI↗

arXiv2024

A Transformer variant for multi-step forecasting of water level and hydrometeorological sensitivity analysis based on explainable artificial intelligence technology

Understanding the combined influences of meteorological and hydrological factors on water level and flood events is essential, particularly in today's changing climate environments. Transformer, as one kind of the cutting-edge deep learning methods, offers an effective approach to model intricate nonlinear processes, enables the extraction of key features and water level predictions. EXplainable Artificial Intelligence (XAI) methods play important roles in enhancing the understandings of how different factors impact water level. In this study, we propose a Transformer variant by integrating sparse attention mechanism and introducing nonlinear output layer for the decoder module. The variant model is utilized for multi-step forecasting of water level, by considering meteorological and hydrological factors simultaneously. It is shown that the variant model outperforms traditional Transformer across different lead times with respect to various evaluation metrics. The sensitivity analyses based on XAI technology demonstrate the significant influence of meteorological factors on water level evolution, in which temperature is shown to be the most dominant meteorological factor. Therefore, incorporating both meteorological and hydrological factors is necessary for reliable hydrological prediction and flood prevention. In the meantime, XAI technology provides insights into certain predictions, which is beneficial for understanding the prediction results and evaluating the reasonability.

cs.LG↗

arXiv2023

Where Did the President Visit Last Week? Detecting Celebrity Trips from News Articles

Celebrities' whereabouts are of pervasive importance. For instance, where politicians go, how often they visit, and who they meet, come with profound geopolitical and economic implications. Although news articles contain travel information of celebrities, it is not possible to perform large-scale and network-wise analysis due to the lack of automatic itinerary detection tools. To design such tools, we have to overcome difficulties from the heterogeneity among news articles: 1)One single article can be noisy, with irrelevant people and locations, especially when the articles are long. 2)Though it may be helpful if we consider multiple articles together to determine a particular trip, the key semantics are still scattered across different articles intertwined with various noises, making it hard to aggregate them effectively. 3)Over 20% of the articles refer to the celebrities' trips indirectly, instead of using the exact celebrity names or location names, leading to large portions of trips escaping regular detecting algorithms. We model text content across articles related to each candidate location as a graph to better associate essential information and cancel out the noises. Besides, we design a special pooling layer based on attention mechanism and node similarity, reducing irrelevant information from longer articles. To make up the missing information resulted from indirect mentions, we construct knowledge sub-graphs for named entities (person, organization, facility, etc.). Specifically, we dynamically update embeddings of event entities like the G7 summit from news descriptions since the properties (date and location) of the event change each time, which is not captured by the pre-trained event representations. The proposed CeleTrip jointly trains these modules, which outperforms all baseline models and achieves 82.53% in the F1 metric.

cs.AI↗

arXiv2023

An Approach to Mismatched Disturbance Rejection Control for Continuous-Time Uncontrollable Systems

This paper focuses on optimal mismatched disturbance rejection control for linear continuoustime uncontrollable systems. Different from previous studies, by introducing a new quadratic performance index to transform the mismatched disturbance rejection control into a linear quadratic tracking problem, the regulated state can track a reference trajectory and minimize the influence of disturbance. The necessary and sufficient conditions for the solvability and the disturbance rejection controller are obtained by solving a forward-backward differential equation over a finite horizon. A sufficient condition for system stability is obtained over an infinite horizon under detectable condition. This paper details our novel approach for transforming disturbance rejection into a linear quadratic tracking problem. The effectiveness of the proposed method is provided with two examples to demonstrate.

math.OC↗

arXiv2023

Boundary Unlearning

The practical needs of the ``right to be forgotten'' and poisoned data removal call for efficient \textit{machine unlearning} techniques, which enable machine learning models to unlearn, or to forget a fraction of training data and its lineage. Recent studies on machine unlearning for deep neural networks (DNNs) attempt to destroy the influence of the forgetting data by scrubbing the model parameters. However, it is prohibitively expensive due to the large dimension of the parameter space. In this paper, we refocus our attention from the parameter space to the decision space of the DNN model, and propose Boundary Unlearning, a rapid yet effective way to unlearn an entire class from a trained DNN model. The key idea is to shift the decision boundary of the original DNN model to imitate the decision behavior of the model retrained from scratch. We develop two novel boundary shift methods, namely Boundary Shrink and Boundary Expanding, both of which can rapidly achieve the utility and privacy guarantees. We extensively evaluate Boundary Unlearning on CIFAR-10 and Vggface2 datasets, and the results show that Boundary Unlearning can effectively forget the forgetting class on image classification and face recognition tasks, with an expected speed-up of $17\times$ and $19\times$, respectively, compared with retraining from the scratch.

cs.CV↗

arXiv2023

For the Underrepresented in Gender Bias Research: Chinese Name Gender Prediction with Heterogeneous Graph Attention Network

Achieving gender equality is an important pillar for humankind's sustainable future. Pioneering data-driven gender bias research is based on large-scale public records such as scientific papers, patents, and company registrations, covering female researchers, inventors and entrepreneurs, and so on. Since gender information is often missing in relevant datasets, studies rely on tools to infer genders from names. However, available open-sourced Chinese gender-guessing tools are not yet suitable for scientific purposes, which may be partially responsible for female Chinese being underrepresented in mainstream gender bias research and affect their universality. Specifically, these tools focus on character-level information while overlooking the fact that the combinations of Chinese characters in multi-character names, as well as the components and pronunciations of characters, convey important messages. As a first effort, we design a Chinese Heterogeneous Graph Attention (CHGAT) model to capture the heterogeneity in component relationships and incorporate the pronunciations of characters. Our model largely surpasses current tools and also outperforms the state-of-the-art algorithm. Last but not least, the most popular Chinese name-gender dataset is single-character based with far less female coverage from an unreliable source, naturally hindering relevant studies. We open-source a more balanced multi-character dataset from an official source together with our code, hoping to help future research promoting gender equality.

cs.AI↗

arXiv2022

Thermal load models for the static design of steel-concrete composite girders

Although the recommended temperature gradient models of composite girders are considered in current specifications for classifying temperature effects in various countries, they are not appropriate enough for static design. Moreover, existing national standards cannot explain the mechanism of the thermal effect on the bridge. To further investigate thermal load models of composite girders, this work proposed a decomposing method for vertical nonlinear temperature gradients based on thermal effects and a calculating method for the thermal stress of composite girders. Equivalent temperature (equivalent uniform temperature, equivalent linear temperature, and equivalent nonlinear temperature), temperature difference, and cyclic equivalent uniform temperature are analyzed to reflect the characteristics of thermal effect in composite girders. The stand values of temperature differences and equivalent temperature with a 50-year return period were investigated via probabilistic statistical analysis. Additionally, two vertical thermal load models (VTLM 1 and VTLM 2) were set up to facilitate the design and applied to stress analysis. The result demonstrates that the proposed thermal load model is more suitable than the Chinese Specification for calculating the thermal effects of composite girders.

physics.app-ph↗

arXiv2022

An Approach to Mismatched Disturbance Rejection Control for Uncontrollable Systems

This study focuses on the problem of optimal mismatched disturbance rejection control for uncontrollable linear discrete-time systems. In contrast to previous studies, by introducing a quadratic performance index such that the regulated state can track a reference trajectory and minimize the effects of disturbances, mismatched disturbance rejection control is transformed into a linear quadratic tracking problem. The necessary and sufficient conditions for the solvability of this problem over a finite horizon and a disturbance rejection controller are derived by solving a forward-backward difference equation. In the case of an infinite horizon, a sufficient condition for the stabilization of the system is obtained under the detectable condition. This paper details our novel approach to disturbance rejection. Four examples are provided to demonstrate the effectiveness of the proposed method.

math.OC↗

arXiv2022

Mismatched Disturbance Rejection Control for Second-Order Discrete-Time Systems

This paper is concerned with mismatched disturbance rejection control for the second-order discrete-time systems.Different from previous work, the controllability of the system is applied to design the disturbance compensation gain, which does not require any coordinate transformations. Via this new idea, it is shown that disturbance in the regulated output is immediately and directly compensated in the case that the disturbance is known. When the disturbance is unknown, an extra generalized extended state observer is applied to design the controller. Two examples are given to show the effectiveness of the proposed methods. Numerical simulation shows that the designed controller has excellent disturbance rejection effect when the disturbance is known. The example with respect to the permanent-magnet direct current motor illustrates that the proposed control method for unknown disturbance rejection is effective.

math.OC↗