arXiv subjects

Sangbum Kim

Publications and source records attributed to Sangbum Kim.

11 recordsLinked to original sources

arXiv2026

Neuronal arithmetic operators based on Ovonic threshold switches (OTS) for biologically inspired analog computing

Biological neurons perform arithmetic computations - including additive integration and divisive gain modulation - through synaptic conductance changes and shunting inhibition, enabling context-dependent information processing that far exceeds simple threshold-and-fire models. Replicating these capabilities in compact hardware remains a fundamental challenge for neuromorphic engineering. Here, we demonstrate artificial neuron circuits based on Ovonic threshold switches (OTS) that physically implement three arithmetic operations: SUM, PARALLEL, and DIVISION. The SUM and PARALLEL neurons exploit MOSFET-controlled dendritic conductances, producing output firing rates that collapse onto invariant curves as a function of combined inputs - satisfying the canonical criteria for neuronal addition. The DIVISION neuron leverages a JFET-based shunting pathway, inspired by GABA_A-mediated inhibition in the cortex, to achieve divisive gain modulation well described by a Hill-type function (R2 ~ 0.95, Hill exponent n ~ 1.3), consistent with nonlinear normalization observed in visual and olfactory circuits. Applying the DIVISION neuron to pixel-wise image normalization under non-uniform illumination recovers obscured visual content, mirroring contrast normalization in the visual cortex. Compared to CMOS-based division implementations, the proposed approach offers improvements in energy efficiency and scalability exceeding an order of magnitude, establishing a viable path toward compact, brain-inspired analog computing.

physics.app-ph↗

arXiv2025

Ovonic switches enable energy-efficient dendrite-like computing

Over the last decade, dendrites within individual biological neurons, which were previously thought to generally perform information pooling and networking, have now been shown to express complex temporal dynamics, Boolean-like logic, arithmetic, signal discrimination, and edge detection for image and sound recognition. Mimicking this rich functional density could offer a powerful primitive for neuromorphic computing, which has sought to replace the aging digital computing paradigms using biological inspirations. Here, using electrically driven Ovonic threshold switching in Sb-Te-doped GeSe, we demonstrate a single two-terminal component capable of self-sustained dynamics and universal Boolean logic, in addition to XOR operations (which is traditionally thought to require a network of active components). We then employ logic-driven dynamics in a single component to detect and estimate the gradients of edges in images, a task that otherwise requires elaborate circuits. A network of Ovonic switches exhibits properties of a half adder and a full adder, in addition to discriminative logic accommodating inhibitory and excitatory signals. We show that this computational primitive is not only seemingly simpler, but also offers many orders of magnitude improved energy efficiency compared to prevailing digital solutions. As such, this work paves the path for potentially emulating dendrites for efficient post-digital neuromorphic computing.

cs.ET↗

arXiv2025

PPMI: Privacy-Preserving LLM Interaction with Socratic Chain-of-Thought Reasoning and Homomorphically Encrypted Vector Databases

Large language models (LLMs) are increasingly used as personal agents, accessing sensitive user data such as calendars, emails, and medical records. Users currently face a trade-off: They can send private records, many of which are stored in remote databases, to powerful but untrusted LLM providers, increasing their exposure risk. Alternatively, they can run less powerful models locally on trusted devices. We bridge this gap. Our Socratic Chain-of-Thought Reasoning first sends a generic, non-private user query to a powerful, untrusted LLM, which generates a Chain-of-Thought (CoT) prompt and detailed sub-queries without accessing user data. Next, we embed these sub-queries and perform encrypted sub-second semantic search using our Homomorphically Encrypted Vector Database across one million entries of a single user's private data. This represents a realistic scale of personal documents, emails, and records accumulated over years of digital activity. Finally, we feed the CoT prompt and the decrypted records to a local language model and generate the final response. On the LoCoMo long-context QA benchmark, our hybrid framework, combining GPT-4o with a local Llama-3.2-1B model, outperforms using GPT-4o alone by up to 7.1 percentage points. This demonstrates a first step toward systems where tasks are decomposed and split between untrusted strong LLMs and weak local ones, preserving user privacy.

cs.CR↗

arXiv2024

CUPID: A Real-Time Session-Based Reciprocal Recommendation System for a One-on-One Social Discovery Platform

This study introduces CUPID, a novel approach to session-based reciprocal recommendation systems designed for a real-time one-on-one social discovery platform. In such platforms, low latency is critical to enhance user experiences. However, conventional session-based approaches struggle with high latency due to the demands of modeling sequential user behavior for each recommendation process. Additionally, given the reciprocal nature of the platform, where users act as items for each other, training recommendation models on large-scale datasets is computationally prohibitive using conventional methods. To address these challenges, CUPID decouples the time-intensive user session modeling from the real-time user matching process to reduce inference time. Furthermore, CUPID employs a two-phase training strategy that separates the training of embedding and prediction layers, significantly reducing the computational burden by decreasing the number of sequential model inferences by several hundredfold. Extensive experiments on large-scale Azar datasets demonstrate CUPID's effectiveness in a real-world production environment. Notably, CUPID reduces response latency by more than 76% compared to non-asynchronous systems, while significantly improving user engagement.

cs.IR↗

arXiv2022

Delocalization and re-entrant localization of flat-band states in non-Hermitian disordered lattice models with flat bands

We present a numerical study of Anderson localization in disordered non-Hermitian lattice models with flat bands. Specifically we consider one-dimensional stub and two-dimensional kagome lattices that have a random scalar potential and a uniform imaginary vector potential and calculate the spectra of the complex energy, the participation ratio, and the winding number as a function of the strength of the imaginary vector potential, $h$. The flat-band states are found to show a double transition from localized to delocalized and back to localized states with $h$, in contrast to the dispersive-band states going through a single delocalization transition. When $h$ is sufficiently small, all flat-band states are localized. As $h$ increases above a certain critical value $h_1$, some pair of flat-band states become delocalized. The participation ratio associated with them increases substantially and their winding numbers become nonzero. As $h$ increases further, more and more flat-band states get delocalized until the fraction of the delocalized states reaches a maximum. For larger $h$ values, a re-entrant localization takes place and, at another critical value $h_2$, all flat-band states return to compact localized states with very small participation ratios and zero winding numbers. This re-entrant localization transition, which is due to the interplay among disorder, non-Hermiticity, and flat band, is a phenomenon occurring in many models having an imaginary vector potential and a flat band simultaneously. We explore the spatial characteristics of the flat-band states by calculating the local density distribution.

cond-mat.dis-nn↗

arXiv2022

Meet Your Favorite Character: Open-domain Chatbot Mimicking Fictional Characters with only a Few Utterances

In this paper, we consider mimicking fictional characters as a promising direction for building engaging conversation models. To this end, we present a new practical task where only a few utterances of each fictional character are available to generate responses mimicking them. Furthermore, we propose a new method named Pseudo Dialog Prompting (PDP) that generates responses by leveraging the power of large-scale language models with prompts containing the target character's utterances. To better reflect the style of the character, PDP builds the prompts in the form of dialog that includes the character's utterances as dialog history. Since only utterances of the characters are available in the proposed task, PDP matches each utterance with an appropriate pseudo-context from a predefined set of context candidates using a retrieval model. Through human and automatic evaluation, we show that PDP generates responses that better reflect the style of fictional characters than baseline methods.

cs.CL↗

arXiv2022

Novel Weight Update Scheme for Hardware Neural Network based on Synaptic Devices Having Abrupt LTP or LTD Characteristics

Mitigating nonlinear weight update characteristics is one of the main challenges in designing neural networks based on synaptic devices. This paper presents a novel weight update method named conditional reverse update scheme (CRUS) for hardware neural network (HNN) consisting of synaptic devices with highly nonlinear or abrupt conductance update characteristics. We formulate a linear optimization method of conductance in synaptic devices to reduce the average deviation of weight changes from those calculated by the Stochastic Gradient Rule (SGD) algorithm. We introduce a metric called update noise (UN) to analyze the training dynamics during training. We then design a weight update rule that reduces the UN averaged over the training process. The optimized network achieves >90% accuracy on the MNIST dataset under highly nonlinear long-term potentiation (LTP) and long-term depression (LTD) conditions while using inaccurate and infrequent conductance sensing. Furthermore, the proposed method shows better accuracy than previously reported nonlinear weight update mitigation techniques under the same hardware specifications and device conditions. It also exhibits robustness to temporal variations in conductance updates. We expect our scheme to relieve design requirements in device and circuit engineering and serve as a practical technique that can be applied to future HNNs.

cs.ET↗

arXiv2019

Direct calculation of the strong Goos-Hänchen effect of a Gaussian light beam due to the excitation of surface plasmon polaritons in the Otto configuration

We study theoretically the influence of the surface plasmon excitation on the Goos-Hänchen lateral shift of a $p$-polarized Gaussian beam incident obliquely on a dielectric-metal bilayer in the Otto configuration. We find that the lateral shift depends sensitively on the thickness of the metal layer and the width of the incident beam, as well as on the incident angle. Near the incident angle at which surface plasmons are excited, the lateral shift changes from large negative values to large positive values as the thickness of the metal layer increases through a critical value. For wide incident beams, the maximal forward and backward lateral shifts can be as large as several hundred times of the wavelength. As the width of the incident Gaussian beam decreases, the magnitude of the lateral shift decreases rapidly, but the ratio of the width of the reflected beam to that of the incident beam, which measures the degree of the deformation of the reflected beam profile, increases. In all cases considered, we find that the reflected beam is split into two parts. We also find that the lateral shift of the transmitted beam is always positive and very weak.

physics.optics↗

arXiv2017

Giant enhancement of reflectance due to the interplay between surface confined wave modes and nonlinear gain in dielectric media

We study theoretically the interplay between the surface confined wave modes and the linear and nonlinear gain of the dielectric layer in the Otto configuration. The surface confined wave modes such as surface plasmons or waveguide modes are excited in the dielectric-metal bilayer by obliquely incident $p$ waves. In the purely linear case, we find that the interplay between linear gain and surface confined wave modes can generate a large reflectance peak with its value much greater than 1. As the linear gain parameter increases, the peak appears at smaller incident angles, and the associated modes also change from surface plasmons to waveguide modes. When the nonlinear gain is turned on, the reflectance shows very strong multistability near the incident angles associated with surface confined wave modes. As the nonlinear gain parameter is varied, the reflectance curve undergoes complicated topological changes and sometimes displays separated closed curves. When the nonlinear gain parameter takes an optimally small value, a giant amplification of the reflectance by three orders of magnitude occurs near the incident angle associated with a waveguide mode. We also find that there exists a range of the incident angle where the wave is dissipated rather than amplified even in the presence of gain. We suggest that this can provide the basis for a possible new technology for thermal control in the subwavelength scale.

physics.optics↗

arXiv2006

Creation and pinning of vortex-antivortex pairs

Computer modeling is reported about the creation and pinning of a magnetic vortex-antivortex (V-AV) pair in a superconducting thin film, due to the magnetic field of a vertical magnetic dipole above the film, and two antidot pins inside the film. For film thickness $= 0.1ξ$, $κ= 2$, and no pins, we find the film carries two V-AV pairs at steady state in the imposed flux range $2.10Φ_0 < Φ^+ < 3.0Φ_0$, and no pairs below. With two antidot pins suitably introduced into the film, a single V-AV pair can be stable in the film for $Φ^+ \ge 1.3Φ_0$. At pin separation $\ge 17ξ$, we find the V-AV pair remains pinned after the dipole field is removed, and, so can represent a 1 for a nonvolatile memory.

cond-mat.supr-con↗

arXiv2003

Steady-state and equilibrium vortex configurations, transitions, and evolution in a mesoscopic superconducting cylinder

A numerical scheme to study the mixed states in a mesoscopic type-II superconducting cylinder is described. Steady-state configurations and transient behavior of the magnetic vortices for various values of the applied magnetic field H are presented. Transitions between different multi-vortex states as H is changed is demonstrated by contour plots and jumps in the B vs H plot. Evolving a uniformly-superconducting initial state using the simplest set of relaxation equations shows that the system passes through nearly metastable intermediate configurations, while seeking the final minimum-energy, steady state consistent with the square symmetry of the sample. An efficient scheme to determine the equilibrium vortex configuration in a mesoscopic system at any given applied field, not limited to the symmetry of the system, is devised and demonstrated.

cond-mat.supr-con↗