arXiv subjects

Junhe Zhang

Publications and source records attributed to Junhe Zhang.

6 recordsLinked to original sources

arXiv2026

Staying In Character: Perspective-Bounded Memory For Book-Based Role-Playing Agents

Recent LLM role-playing systems build character agents from novels by extracting characters, scenes, and relations. Yet long-narrative role-playing suffers from two failures: Factual Overreach, where shared retrieval or parametric memory lets a character use facts outside its perspective, and Stylistic Monotony, where profile descriptions flatten a character into a fixed voice. To address these failures, we propose REVERIEMEM, a three-layer memory architecture for book-based character agents. The episodic layer stores first-person scene memories; the semantic layer stores visibility-tagged facts; and the personality layer stores situation-dependent speech and behaviour patterns. For evaluation, we construct KBF-QA, a 4,386-question benchmark over eight novels for testing knowledge boundaries. REVERIEMEM improves Knowledge Boundary Fidelity by 34.6 percentage points over the strongest prior method. On BOOKWORLD's five-dimension pairwise narrative protocol, REVERIEMEM achieves a ~ 79% win rate, suggesting that perspective-bounded memory improves both boundary fidelity and character-grounded narrative generation.

cs.CL↗

arXiv2025

Electrically driven plasmon-polaritonic bistability in Dirac electron tunneling transistors

Bistability-two distinct stable states under identical parameter-is not only a fundamental physical concept but also of importance in practical applications. While plasmon-polaritonic bistability representing history-dependent stable states within plasmonic systems has been theoretically predicted, it has yet to be demonstrated experimentally due to challenges in realizing suitable nonlinearity at feasible electric-field strengths. Here, we report the experimental observation of electrically driven plasmon-polaritonic bistability in graphene/hexagonal-boron-nitride/graphene tunneling transistors, achieved through momentum-conserving resonant tunneling of Dirac electrons. Using a small twist angle between graphene layers, we engineered devices exhibiting both electronic and plasmon-polaritonic bistability. This bistable plasmonic behavior can be precisely tuned through load resistance and electrostatic gating. Our findings open new pathways for exploring nonlinear optical and electronic phenomena in van der Waals heterostructures and mark a significant advance in nanoplasmonics, with potential applications in optical memory, sensing, and optoelectronic switching.

cond-mat.mes-hall↗

arXiv2025

Cavity-altered superconductivity

Is it feasible to alter the ground state properties of a material by engineering its electromagnetic environment? Inspired by theoretical predictions, experimental realizations of such cavity-controlled properties without optical excitation are beginning to emerge. Here, we devised and implemented a novel platform to realize cavity-altered materials. Single crystals of hyperbolic van der Waals (vdW) compounds provide a resonant electromagnetic environment with enhanced density of photonic states and prominent mode confinement. We interfaced hexagonal boron nitride (hBN) with the molecular superconductor $\kappa$-(BEDT-TTF)$_2$Cu[N(CN)$_2$]Br ($\kappa$-ET). The frequencies of infrared (IR) hyperbolic modes of hBN match the IR-active carbon-carbon stretching molecular resonance of ($\kappa$-ET) implicated in superconductivity. Nano-optical data supported by first-principles molecular Langevin dynamics simulations confirm the presence of resonant coupling between the hBN hyperbolic cavity modes and the carbon-carbon stretching mode in ($\kappa$-ET). Meissner effect measurements via magnetic force microscopy demonstrate a strong suppression of superfluid density near the hBN/($\kappa$-ET) interface. Non-resonant control heterostructures, including RuCl$_3$/($\kappa$-ET) and hBN/$\text{Bi}_2\text{Sr}_2\text{CaCu}_2\text{O}_{8+x}$, do not display the superfluid suppression. These observations suggest that hBN/($\kappa$-ET) realizes a cavity-altered superconducting ground state. Our work highlights the potential of dark cavities devoid of external photons for engineering electronic ground state properties of complex quantum materials.

cond-mat.supr-con↗

arXiv2025

Task-Oriented Multimodal Token Transmission in Resource-Constrained Multiuser Networks

With the emergence of large model-based agents, widely adopted transformer-based architectures inevitably produce excessively long token embeddings for transmission, which may result in high bandwidth overhead, increased power consumption and latency. In this letter, we propose a task-oriented multimodal token transmission scheme for efficient multimodal information fusion and utilization. To improve the efficiency of token transmission, we design a two-stage training algotithm, including cross-modal alignment and task-oriented fine-tuning, for large model-based token communication. Meanwhile, token compression is performed using a sliding window pooling operation to save communication resources. To balance the trade-off between latency and model performance caused by compression, we formulate a weighted-sum optimization problem over latency and validation loss. We jointly optimizes bandwidth, power allocation, and token length across users by using an alternating optimization method. Simulation results demonstrate that the proposed algorithm outperforms the baseline under different bandwidth and power budgets. Moreover, the two-stage training algorithm achieves higher accuracy across various signal-to-noise ratios than the method without cross-modal alignment.

cs.NI↗

arXiv2024

Federated Split Learning with Model Pruning and Gradient Quantization in Wireless Networks

As a paradigm of distributed machine learning, federated learning typically requires all edge devices to train a complete model locally. However, with the increasing scale of artificial intelligence models, the limited resources on edge devices often become a bottleneck for efficient fine-tuning. To address this challenge, federated split learning (FedSL) implements collaborative training across the edge devices and the server through model splitting. In this paper, we propose a lightweight FedSL scheme, that further alleviates the training burden on resource-constrained edge devices by pruning the client-side model dynamicly and using quantized gradient updates to reduce computation overhead. Additionally, we apply random dropout to the activation values at the split layer to reduce communication overhead. We conduct theoretical analysis to quantify the convergence performance of the proposed scheme. Finally, simulation results verify the effectiveness and advantages of the proposed lightweight FedSL in wireless network environments.

cs.LG↗

arXiv2024

Plasmonic polarization sensing of electrostatic superlattice potentials

Plasmon polaritons are formed by coupling light with delocalized electrons. The half-light and half-matter nature of plasmon polaritons endows them with unparalleled tunability via a range of parameters, such as dielectric environments and carrier density. Therefore, plasmon polaritons are expected to be tuned when in proximity to polar materials since the carrier density is tuned by an electrostatic potential; conversely, the plasmon polariton response might enable the sensing of polarization. Here, we use infrared nano-imaging and nano-photocurrent measurements to investigate heterostructures composed of graphene and twisted hexagonal boron nitride (t-BN), with alternating polarization in a triangular network of moir\'e stacking domains. We observe that the carrier density and the corresponding plasmonic response of graphene are modulated by polar domains in t-BN. In addition, we demonstrate that the nanometer-wide domain walls of graphene moir\'es superlattices, created by the polar domains of t-BN, provide momenta to assist the plasmonic excitations. Furthermore, our studies establish that the plasmon of graphene could function as a delicate sensor for polarization textures. The evolution of polarization textures in t-BN under uniform electric fields is tomographically examined via plasmonic imaging. Strikingly, no noticeable polarization switching is observed under applied electric fields up to 0.23 V/nm, at variance with transport reports. Our nano-images unambiguously reveal that t-BN with triangular domains acts like a ferrielectric, rather than ferroelectric claimed by many previous studies.

cond-mat.mes-hall↗