arXiv subjects

Zhiming Wang

Publications and source records attributed to Zhiming Wang.

At least 19 recordsLinked to original sources

arXiv2026

A Diamagnetic, Light-Driven Tesla Engine Based on a Mechanically Displaced, Magnetically Levitated Graphene Disk

Ferromagnetic materials are widely used in Tesla thermomagnetic engines, whereas diamagnetic counterparts have remained unexplored. Here, we demonstrate the first diamagnetic Tesla engine by exploiting the strong diamagnetism of graphene. A graphene disk, fabricated by stacking graphene sheets, serves as the engine wheel. We first show that the conventional Tesla engine design using a permanent magnet placed near the disk edge to create unbalanced thermomagnetic forces under asymmetric local heating fails to generate rotation. We achieve stable operation by laterally displacing the levitated disk from equilibrium, creating a strong restoring force that drives rotation under light excitation. Calculations and measurements establish the displacement-dependent force, with an optimal offset of 0.8 mm yielding speeds up to 2000 rpm under laser heating and 1000 rpm under direct sunlight. Adding vanes allows the disk to function as a gear, powering a graphene vehicle and transferring energy to another disk. This design utilizes the strong and anisotropic diamagnetism of graphene and paves the way for light-powered sensors, actuators, and micro-vehicles.

physics.app-ph↗

arXiv2026

Beyond Token Eviction: Mixed-Dimension Budget Allocation for Efficient KV Cache Compression

Key-value (KV) caching is widely used to accelerate transformer inference, but its memory cost grows linearly with input length, limiting long-context deployment. Existing token eviction methods reduce memory by discarding less important tokens, which can be viewed as a coarse form of dimensionality reduction that assigns each token either zero or full dimension. We propose MixedDimKV, a mixed-dimension KV cache compression method that allocates dimensions to tokens at a more granular level, and MixedDimKV-H, which further integrates head-level importance information. Experiments on long-context benchmarks show that MixedDimKV outperforms prior KV cache compression methods that do not rely on head-level importance profiling. When equipped with the same head-level importance information, MixedDimKV-H consistently outperforms HeadKV. Notably, our approach achieves comparable performance to full attention on LongBench with only 6.25% of the KV cache. Furthermore, in the Needle-in-a-Haystack test, our solution maintains 100% accuracy at a 50K context length while using as little as 0.26% of the cache.

cs.LG↗

arXiv2026

Requesting Expert Reasoning: Augmenting LLM Agents with Learned Collaborative Intervention

Large Language Model (LLM) based agents excel at general reasoning but often fail in specialized domains where success hinges on long-tail knowledge absent from their training data. While human experts can provide this missing knowledge, their guidance is often unstructured and unreliable, making its direct integration into an agent's plan problematic. To address this, we introduce AHCE (Active Human-Augmented Challenge Engagement), a framework for on-demand Human-AI collaboration. At its core, the Human Feedback Module (HFM) employs a learned policy to treat the human expert as an interactive reasoning tool. Extensive experiments in Minecraft demonstrate the framework's effectiveness, increasing task success rates by 32% on normal difficulty tasks and nearly 70% on highly difficult tasks, all with minimal human intervention. Our work demonstrates that successfully augmenting agents requires learning how to request expert reasoning, moving beyond simple requests for help.

cs.AI↗

arXiv2026

Mechanisms and Opportunities for Tunable High-Purity Single Photon Emitters: A Review of Hybrid Perovskites and Prospects for Bright Squeezed Vacuum

Single-photon emitters (SPEs) are central to quantum communication, computing, and metrology, yet their development remains constrained by trade-offs in purity, indistinguishability, and tunability. This review presents a mechanism-based classification of SPEs, offering a physics-oriented framework to clarify the performance limitations of conventional sources, including quantum emitters and nonlinear optical processes. Particular attention is given to hybrid organic-inorganic perovskite quantum dots (HOIP QDs), which provide size- and composition-tunable emission with narrow linewidths and room-temperature operation. Through comparative analysis of physical mechanisms and performance metrics, we show how HOIP QDs may address key limitations of established SPE platforms. Recognizing the constraints of current deterministic sources, we introduce a performance framework to guide the development of scalable SPEs, and examine the theoretical potential of bright squeezed vacuum (BSV) states, discussing how BSV mechanisms could serve as a promising avenue for multiplexable, high-purity photon generation beyond conventional heralded schemes. The review concludes by outlining future directions for integrating HOIP- and BSV-based concepts into scalable quantum photonic architectures.

physics.optics↗

arXiv2025

KVReviver: Reversible KV Cache Compression with Sketch-Based Token Reconstruction

As the context length of current large language models (LLMs) rapidly increases, the memory demand for the Key-Value (KV) cache is becoming a bottleneck for LLM deployment and batch processing. Traditional KV cache compression methods typically involve permanently evicting or irreversibly merging "less important" tokens with low attention scores. This approach results in the unrecoverable loss of token information, which we call Contextual Amnesia, significantly degrading the model's information retrieval capability. To address this issue, we propose KVReviver, a reversible KV cache compression method based on the sketch algorithm. This method allows reconstructing compressed tokens from an additional data structure, thus enabling full-scale computation within limited memory. Experiments showed that in 2k-length contexts, it requires only 10% of KV Cache budget while maintaining identical end-to-end inference accuracy. For 32k-length contexts, it achieves equivalent or comparable accuracy ~2% accuracy loss) using merely 25% of KV Cache budget.

cs.CL↗

arXiv2025

PT-DETR: Small Target Detection Based on Partially-Aware Detail Focus

To address the challenges in UAV object detection, such as complex backgrounds, severe occlusion, dense small objects, and varying lighting conditions,this paper proposes PT-DETR based on RT-DETR, a novel detection algorithm specifically designed for small objects in UAV imagery. In the backbone network, we introduce the Partially-Aware Detail Focus (PADF) Module to enhance feature extraction for small objects. Additionally,we design the Median-Frequency Feature Fusion (MFFF) module,which effectively improves the model's ability to capture small-object details and contextual information. Furthermore,we incorporate Focaler-SIoU to strengthen the model's bounding box matching capability and increase its sensitivity to small-object features, thereby further enhancing detection accuracy and robustness. Compared with RT-DETR, our PT-DETR achieves mAP improvements of 1.6% and 1.7% on the VisDrone2019 dataset with lower computational complexity and fewer parameters, demonstrating its robustness and feasibility for small-object detection tasks.

cs.CV↗

arXiv2025

Prototype-Driven Adaptation for Few-Shot Object Detection

Few-shot object detection (FSOD) often suffers from base-class bias and unstable calibration when only a few novel samples are available. We propose Prototype-Driven Alignment (PDA), a lightweight, plug-in metric head for DeFRCN that provides a prototype-based "second opinion" complementary to the linear classifier. PDA maintains support-only prototypes in a learnable identity-initialized projection space and optionally applies prototype-conditioned RoI alignment to reduce geometric mismatch. During fine-tuning, prototypes can be adapted via exponential moving average(EMA) updates on labeled foreground RoIs-without introducing class-specific parameters-and are frozen at inference to ensure strict protocol compliance. PDA employs a best-of-K matching scheme to capture intra-class multi-modality and temperature-scaled fusion to combine metric similarities with detector logits. Experiments on VOC FSOD and GFSOD benchmarks show that PDA consistently improves novel-class performance with minimal impact on base classes and negligible computational overhead.

cs.CV↗

arXiv2025

3rd Place Solution to ICCV LargeFineFoodAI Retrieval

This paper introduces the 3rd place solution to the ICCV LargeFineFoodAI Retrieval Competition on Kaggle. Four basic models are independently trained with the weighted sum of ArcFace and Circle loss, then TTA and Ensemble are successively applied to improve feature representation ability. In addition, a new reranking method for retrieval is proposed based on diffusion and k-reciprocal reranking. Finally, our method scored 0.81219 and 0.81191 mAP@100 on the public and private leaderboard, respectively.

cs.CV↗

arXiv2025

MergeMoE: Efficient Compression of MoE Models via Expert Output Merging

The Mixture-of-Experts (MoE) technique has proven to be a promising solution to efficiently scale the model size, which has been widely applied in recent LLM advancements. However, the substantial memory overhead of MoE models has made their compression an important research direction. In this work, we provide a theoretical analysis of expert merging, a recently proposed technique for compressing MoE models. Rather than interpreting expert merging from the conventional perspective of parameter aggregation, we approach it from the perspective of merging experts' outputs. Our key insight is that the merging process can be interpreted as inserting additional matrices into the forward computation, which naturally leads to an optimization formulation. Building on this analysis, we introduce MergeMoE, a method that leverages mathematical optimization to construct the compression matrices. We evaluate MergeMoE on multiple MoE models and show that our algorithm consistently outperforms the baselines with the same compression ratios.

cs.LG↗

arXiv2025

Flexible orbital torque device with ultralow switching current

Orbital torque (OT) offers a highly efficient way for electrical magnetization manipulation. However, its potential in the emerging field of flexible spintronics remains largely unexplored. Here, we demonstrate a flexible and robust OT device based on a mica/SrRuO3(SRO)/CoPt heterostructure. We measure a large torque efficiency of -0.31, which originates from the significant orbital Hall effect in the SRO layer. Leveraging the low thermal conductivity of the mica substrate, a thermally-assisted switching mechanism is activated, enabling an ultralow threshold current density of 9.2x109 A/m2. This value represents a 90% reduction compared to conventional spin-torque devices and a 52% reduction against its rigid counterpart on a SrTiO3 substrate. The superior performances is well-maintained after 103 bending cycles, conforming its exceptional flexibility and durability. Our work pioneers the development of flexible OT devices, showcasing a viable path toward next-generation, low-power wearable spintronic applications.

cond-mat.mtrl-sci↗

arXiv2025

Intrinsic Strain-Driven Topological Evolution in SrRuO3 via Flexural Strain Engineering

Strain engineering offers a powerful route to tailor topological electronic structures in correlated oxides, yet conventional epitaxial strain approaches introduce extrinsic factors such as substrate-induced phase transitions and crystalline quality variations, which makes the unambiguous identification of the intrinsic strain effects challenging. Here, we develop a flexural strain platform based on van der Waals epitaxy and flexible micro-fabrication, enabling precise isolation and quantification of intrinsic strain effects on topological electronic structures in correlated oxides without extrinsic interference. Through strain-dependent transport measurements of the Weyl semimetal SrRuO3, we observed a significant enhancement of anomalous Hall conductivity by 21% under a tiny strain level of 0.2%, while longitudinal resistivity remains almost constant -- a hallmark of intrinsic topological response. First-principles calculations reveal a distinct mechanism where strain-driven non-monotonic evolution of Weyl nodes across the Fermi level, exclusively governed by lattice constant modulation, drives the striking AHC behavior. Our work not only highlights the pivotal role of pure lattice strain in topological regulation but also establishes a universal platform for designing flexible topological oxide devices with tailored functionalities.

cond-mat.mtrl-sci↗

arXiv2025

Generalizable Audio Deepfake Detection via Hierarchical Structure Learning and Feature Whitening in Poincar\'e sphere

Audio deepfake detection (ADD) faces critical generalization challenges due to diverse real-world spoofing attacks and domain variations. However, existing methods primarily rely on Euclidean distances, failing to adequately capture the intrinsic hierarchical structures associated with attack categories and domain factors. To address these issues, we design a novel framework Poin-HierNet to construct domain-invariant hierarchical representations in the Poincar\'e sphere. Poin-HierNet includes three key components: 1) Poincar\'e Prototype Learning (PPL) with several data prototypes aligning sample features and capturing multilevel hierarchies beyond human labels; 2) Hierarchical Structure Learning (HSL) leverages top prototypes to establish a tree-like hierarchical structure from data prototypes; and 3) Poincar\'e Feature Whitening (PFW) enhances domain invariance by applying feature whitening to suppress domain-sensitive features. We evaluate our approach on four datasets: ASVspoof 2019 LA, ASVspoof 2021 LA, ASVspoof 2021 DF, and In-The-Wild. Experimental results demonstrate that Poin-HierNet exceeds state-of-the-art methods in Equal Error Rate.

cs.SD↗

arXiv2025

SpeechFake: A Large-Scale Multilingual Speech Deepfake Dataset Incorporating Cutting-Edge Generation Methods

As speech generation technology advances, the risk of misuse through deepfake audio has become a pressing concern, which underscores the critical need for robust detection systems. However, many existing speech deepfake datasets are limited in scale and diversity, making it challenging to train models that can generalize well to unseen deepfakes. To address these gaps, we introduce SpeechFake, a large-scale dataset designed specifically for speech deepfake detection. SpeechFake includes over 3 million deepfake samples, totaling more than 3,000 hours of audio, generated using 40 different speech synthesis tools. The dataset encompasses a wide range of generation techniques, including text-to-speech, voice conversion, and neural vocoder, incorporating the latest cutting-edge methods. It also provides multilingual support, spanning 46 languages. In this paper, we offer a detailed overview of the dataset's creation, composition, and statistics. We also present baseline results by training detection models on SpeechFake, demonstrating strong performance on both its own test sets and various unseen test sets. Additionally, we conduct experiments to rigorously explore how generation methods, language diversity, and speaker variation affect detection performance. We believe SpeechFake will be a valuable resource for advancing speech deepfake detection and developing more robust models for evolving generation techniques.

cs.SD↗

arXiv2025

BR-ASR: Efficient and Scalable Bias Retrieval Framework for Contextual Biasing ASR in Speech LLM

While speech large language models (SpeechLLMs) have advanced standard automatic speech recognition (ASR), contextual biasing for named entities and rare words remains challenging, especially at scale. To address this, we propose BR-ASR: a Bias Retrieval framework for large-scale contextual biasing (up to 200k entries) via two innovations: (1) speech-and-bias contrastive learning to retrieve semantically relevant candidates; (2) dynamic curriculum learning that mitigates homophone confusion which negatively impacts the final performance. The is a general framework that allows seamless integration of the retrieved candidates into diverse ASR systems without fine-tuning. Experiments on LibriSpeech test-clean/-other achieve state-of-the-art (SOTA) biased word error rates (B-WER) of 2.8%/7.1% with 2000 bias words, delivering 45% relative improvement over prior methods. BR-ASR also demonstrates high scalability: when expanding the bias list to 200k where traditional methods generally fail, it induces only 0.3 / 2.9% absolute WER / B-WER degradation with a 99.99% pruning rate and only 20ms latency per query on test-other.

cs.SD↗

arXiv2025

KeepKV: Achieving Periodic Lossless KV Cache Compression for Efficient LLM Inference

Efficient inference of large language models (LLMs) is hindered by an ever-growing key-value (KV) cache, making KV cache compression a critical research direction. Traditional methods selectively evict less important KV cache entries, which leads to information loss and hallucinations. Recently, merging-based strategies have been explored to retain more information by merging KV pairs that would be discarded; however, these existing approaches inevitably introduce inconsistencies in attention distributions before and after merging, causing degraded generation quality. To overcome this challenge, we propose KeepKV, a novel adaptive KV cache merging method designed to preserve performance under strict memory constraints, achieving single-step lossless compression and providing error bounds for multi-step compression. KeepKV introduces the Electoral Votes mechanism that records merging history and adaptively adjusts attention scores. Moreover, it further leverages a novel Zero Inference-Perturbation Merging method, compensating for attention loss resulting from cache merging. Extensive experiments on various benchmarks and LLM architectures demonstrate that KeepKV substantially reduces memory usage while successfully retaining essential context information, achieving over 2x inference throughput improvement and maintaining superior generation quality even with only 10% KV cache budgets.

cs.LG↗

arXiv2025

Plasmon-interband hybridization and anomalous production of hot electrons in aluminum nanoantennas

Strong coupling typically occurs between two separate objects or between an object and its environment (such as an atom and a cavity). However, it can also occur between two different excitations within the same object, a situation that has been much less studied. In this study, we observe strong coupling between localized surface plasmon resonances and the interband transition in aluminum nanorods, as evidenced by optical spectroscopy and electron energy loss spectroscopy, and corroborated with numerical simulations. Strong coupling is observed between the interband transition and multiple orders of the surface plasmon mode, including dark ones. We also obtain experimental maps of the hybrid modes at the nanoscale. In each case, the associated Rabi energy, which corresponds to the energy splitting between the two polaritonic branches, is obtained. Finally, a dedicated numerical model was employed to calculate the hot electron generation rate in the nanorods. The calculations demonstrate that efficient generation of hot electrons can be achieved in the near-infrared region, when the interband transition is strongly coupled with a plasmon resonance. This high generation rate stems from the hybrid nature of the mode, as its plasmonic component provides a high absorption cross-section, while the IT part ensures efficient conversion to hot electrons. Consequently, aluminum nanorods represent an efficient source of hot electrons in the visible and near-infrared regions, with potential applications in local photochemistry, photodetection, and solar energy harvesting.

physics.optics↗

arXiv2025

Insulator-to-Metal Transition via Magnetic Reconstruction at Oxide Interfaces

Ultrathin two-dimensional (2D) electronic systems at the interfaces of layered materials are highly desirable platforms for exploring of novel quantum phenomena and developing advanced device applications. Here, we investigate ultrathin heterostructures composed of SrIrO3 (SIO) and SrRuO3 (SRO) layers to uncover their emergent properties. Strikingly, despite the fact that both individual layers are antiferromagnetic insulators, the interfaced heterostructure exhibits emergent metallicity. Through transport measurements, magnetic characterization, and angle-resolved photoemission spectroscopy (ARPES), we analyze the underlying mechanisms governing this insulator-to-metal transition. Our findings reveal that the transition is driven by interface-induced magnetic reconstruction, which is further corroborated by density functional theory (DFT) calculations. The staggered Dzyaloshinskii-Moriya interaction at the SIO/SRO interface is identified as the key driving force for this spin reorganization, as it stabilizes ferromagnetism in the coupled antiferromagnetic insulating layers. These findings highlight the significant potential of engineering interfacial magnetic interactions as a powerful approach to generate and control emergent electronic properties, paving the way for novel functionalities that are unattainable in individual ultrathin films.

cond-mat.mes-hall↗

arXiv2025

Nonlinear Chiroptical Effect in Refractory Plasmonic Molybdenum Metasurface: Chiral Saturated Absorption and Chiral Reverse Saturable Absorption

The absorptive chiral metasurface shows evidence of giant nonlinear chiroptical effects, including chiral third harmonic generation, chiral saturated absorption, and chiral reverse saturable absorption. We finally demonstrate proof of concept circular polarized light limiter based on the refractory chiral Mo metasurface and paves the way for nonlinear chiral sensing, nonlinear hot electron generation, and XOR gate of the all-optical full-adder, showcasing the potential of refractory chiral Mo metasurfaces in overcoming the limitations of traditional chiral plasmonic materials.

physics.optics↗