arXiv subjects

Hyunwoo Oh

Publications and source records attributed to Hyunwoo Oh.

At least 37 records · Page 2Linked to original sources

arXiv2026

Follow the Saliency: Supervised Saliency for Retrieval-augmented Dense Video Captioning

Existing retrieval-augmented approaches for Dense Video Captioning (DVC) often fail to achieve accurate temporal segmentation aligned with true event boundaries, as they rely on heuristic strategies that overlook ground truth event boundaries. The proposed framework, \textbf{STaRC}, overcomes this limitation by supervising frame-level saliency through a highlight detection module. Note that the highlight detection module is trained on binary labels derived directly from DVC ground truth annotations without the need for additional annotation. We also propose to utilize the saliency scores as a unified temporal signal that drives retrieval via saliency-guided segmentation and informs caption generation through explicit Saliency Prompts injected into the decoder. By enforcing saliency-constrained segmentation, our method produces temporally coherent segments that align closely with actual event transitions, leading to more accurate retrieval and contextually grounded caption generation. We conduct comprehensive evaluations on the YouCook2 and ViTT benchmarks, where STaRC achieves state-of-the-art performance across most of the metrics. Our code is available at https://github.com/ermitaju1/STaRC

cs.CV↗

arXiv2026

$n$-Musketeers: Reinforcement Learning Shapes Collaboration Among Language Models

Recent progress in reinforcement learning with verifiable rewards (RLVR) shows that small, specialized language models (SLMs) can exhibit structured reasoning without relying on large monolithic LLMs. We introduce soft hidden-state collaboration, where multiple heterogeneous frozen SLM experts are integrated through their internal representations via a trainable attention interface. Experiments on Reasoning Gym and GSM8K show that this latent integration is competitive with strong single-model RLVR baselines. Ablations further reveal a dual mechanism of expert utilization: for simpler arithmetic domains, performance gains can largely be explained by static expert preferences, whereas more challenging settings induce increasingly concentrated and structured expert attention over training, indicating emergent specialization in how the router connects to relevant experts. Overall, hidden-state collaboration provides a compact mechanism for leveraging frozen experts, while offering an observational window into expert utilization patterns and their evolution under RLVR.

cs.LG↗

arXiv2026

DecoHD: Decomposed Hyperdimensional Classification under Extreme Memory Budgets

Decomposition is a proven way to shrink deep networks without changing input-output dimensionality or interface semantics. We bring this idea to hyperdimensional computing (HDC), where footprint cuts usually shrink the feature axis and erode concentration and robustness. Prior HDC decompositions decode via fixed atomic hypervectors, which are ill-suited for compressing learned class prototypes. We introduce DecoHD, which learns directly in a decomposed HDC parameterization: a small, shared set of per-layer channels with multiplicative binding across layers and bundling at the end, yielding a large representational space from compact factors. DecoHD compresses along the class axis via a lightweight bundling head while preserving native bind-bundle-score; training is end-to-end, and inference remains pure HDC, aligning with in/near-memory accelerators. In evaluation, DecoHD attains extreme memory savings with only minor accuracy degradation under tight deployment budgets. On average it stays within about 0.1-0.15% of a strong non-reduced HDC baseline (worst case 5.7%), is more robust to random bit-flip noise, reaches its accuracy plateau with up to ~97% fewer trainable parameters, and--in hardware--delivers roughly 277x/35x energy/speed gains over a CPU (AMD Ryzen 9 9950X), 13.5x/3.7x over a GPU (NVIDIA RTX 4090), and 2.0x/2.4x over a baseline HDC ASIC.

cs.LG↗

arXiv2026

LogHD: Robust Compression of Hyperdimensional Classifiers via Logarithmic Class-Axis Reduction

Hyperdimensional computing (HDC) suits memory, energy, and reliability-constrained systems, yet the standard "one prototype per class" design requires $O(CD)$ memory (with $C$ classes and dimensionality $D$). Prior compaction reduces $D$ (feature axis), improving storage/compute but weakening robustness. We introduce LogHD, a logarithmic class-axis reduction that replaces the $C$ per-class prototypes with $n\!\approx\!\lceil\log_k C\rceil$ bundle hypervectors (alphabet size $k$) and decodes in an $n$-dimensional activation space, cutting memory to $O(D\log_k C)$ while preserving $D$. LogHD uses a capacity-aware codebook and profile-based decoding, and composes with feature-axis sparsification. Across datasets and injected bit flips, LogHD attains competitive accuracy with smaller models and higher resilience at matched memory. Under equal memory, it sustains target accuracy at roughly $2.5$-$3.0\times$ higher bit-flip rates than feature-axis compression; an ASIC instantiation delivers $498\times$ energy efficiency and $62.6\times$ speedup over an AMD Ryzen 9 9950X and $24.3\times$/$6.58\times$ over an NVIDIA RTX 4090, and is $4.06\times$ more energy-efficient and $2.19\times$ faster than a feature-axis HDC ASIC baseline.

cs.LG↗

arXiv2026

HYPERDOA: Robust and Efficient DoA Estimation using Hyperdimensional Computing

Direction of Arrival (DoA) estimation techniques face a critical trade-off, as classical methods often lack accuracy in challenging, low signal-to-noise ratio (SNR) conditions, while modern deep learning approaches are too energy-intensive and opaque for resource-constrained, safety-critical systems. We introduce HYPERDOA, a novel estimator leveraging Hyperdimensional Computing (HDC). The framework introduces two distinct feature extraction strategies -- Mean Spatial-Lag Autocorrelation and Spatial Smoothing -- for its HDC pipeline, and then reframes DoA estimation as a pattern recognition problem. This approach leverages HDC's inherent robustness to noise and its transparent algebraic operations to bypass the expensive matrix decompositions and "black-box" nature of classical and deep learning methods, respectively. Our evaluation demonstrates that HYPERDOA achieves ~35.39% higher accuracy than state-of-the-art methods in low-SNR, coherent-source scenarios. Crucially, it also consumes ~93% less energy than competing neural baselines on an embedded NVIDIA Jetson Xavier NX platform. This dual advantage in accuracy and efficiency establishes HYPERDOA as a robust and viable solution for mission-critical applications on edge devices.

eess.SP↗

arXiv2026

Numerical study of computational cost of maintaining adiabaticity for long paths

Recent work argued that the scaling of a dimensionless quantity $Q_D$ with path length is a better proxy for quantifying the scaling of the computational cost of maintaining adiabaticity than the timescale. It also conjectured that generically the scaling will be superlinear (although special cases exist in which it is linear). The quantity $Q_D$ depends only on the properties of ground states along the Hamiltonian path and the rate at which the path is followed. In this paper, we demonstrate that this conjecture holds for simple Hamiltonian systems that can be studied numerically. In particular, the systems studied exhibit the behavior that $Q_D$ grows approximately as $L \log L$ where $L$ is the path length when the threshold error is fixed.

quant-ph↗

arXiv2025

Practical limitations of the switching theorem for adiabatic state preparation

The viability of adiabatic quantum computation depends on the slow evolution of the Hamiltonian. The adiabatic switching theorem provides an asymptotic series for error estimates in $1/T$, based on the lowest non-zero derivative of the Hamiltonian and its eigenvalues at the endpoints. Modifications at the endpoints in practical implementations can modify this scaling behavior, suggesting opportunities for error reduction by altering endpoint behavior while keeping intermediate evolution largely unchanged. Such modifications can significantly reduce errors for long evolution times, but they may also require exceedingly long timescales to reach the hyperadiabatic regime, limiting their practicality. This paper explores the transition between the adiabatic and hyperadiabatic regimes in simple low-dimensional Hamiltonians, highlighting the impact of modifications of the endpoints on approaching the asymptotic behavior described by the switching theorem.

quant-ph↗

arXiv2025

T-SAR: A Full-Stack Co-design for CPU-Only Ternary LLM Inference via In-Place SIMD ALU Reorganization

Recent advances in LLMs have outpaced the computational and memory capacities of edge platforms that primarily employ CPUs, thereby challenging efficient and scalable deployment. While ternary quantization enables significant resource savings, existing CPU solutions rely heavily on memory-based lookup tables (LUTs) which limit scalability, and FPGA or GPU accelerators remain impractical for edge use. This paper presents T-SAR, the first framework to achieve scalable ternary LLM inference on CPUs by repurposing the SIMD register file for dynamic, in-register LUT generation with minimal hardware modifications. T-SAR eliminates memory bottlenecks and maximizes data-level parallelism, delivering 5.6-24.5x and 1.1-86.2x improvements in GEMM latency and GEMV throughput, respectively, with only 3.2% power and 1.4% area overheads in SIMD units. T-SAR achieves up to 2.5-4.9x the energy efficiency of an NVIDIA Jetson AGX Orin, establishing a practical approach for efficient LLM inference on edge platforms.

cs.AR↗

arXiv2025

QUILL: An Algorithm-Architecture Co-Design for Cache-Local Deformable Attention

Deformable transformers deliver state-of-the-art detection but map poorly to hardware due to irregular memory access and low arithmetic intensity. We introduce QUILL, a schedule-aware accelerator that turns deformable attention into cache-friendly, single-pass work. At its core, Distance-based Out-of-Order Querying (DOOQ) orders queries by spatial proximity; the look-ahead drives a region prefetch into an alternate buffer--forming a schedule-aware prefetch loop that overlaps memory and compute. A fused MSDeformAttn engine executes interpolation, Softmax, aggregation, and the final projection (W''m) in one pass without spilling intermediates, while small tensors are kept on-chip and surrounding dense layers run on integrated GEMMs. Implemented as RTL and evaluated end-to-end, QUILL achieves up to 7.29x higher throughput and 47.3x better energy efficiency than an RTX 4090, and exceeds prior accelerators by 3.26-9.82x in throughput and 2.01-6.07x in energy efficiency. With mixed-precision quantization, accuracy tracks FP32 within <=0.9 AP across Deformable and Sparse DETR variants. By converting sparsity into locality--and locality into utilization--QUILL delivers consistent, end-to-end speedups.

cs.AR↗

arXiv2025

ASTER: Attention-based Spiking Transformer Engine for Event-driven Reasoning

The integration of spiking neural networks (SNNs) with transformer-based architectures has opened new opportunities for bio-inspired low-power, event-driven visual reasoning on edge devices. However, the high temporal resolution and binary nature of spike-driven computation introduce architectural mismatches with conventional digital hardware (CPU/GPU). Prior neuromorphic and Processing-in-Memory (PIM) accelerators struggle with high sparsity and complex operations prevalent in such models. To address these challenges, we propose a memory-centric hardware accelerator tailored for spiking transformers, optimized for deployment in real-time event-driven frameworks such as classification with both static and event-based input frames. Our design leverages a hybrid analog-digital PIM architecture with input sparsity optimizations, and a custom-designed dataflow to minimize memory access overhead and maximize data reuse under spatiotemporal sparsity, for compute and memory-efficient end-to-end execution of spiking transformers. We subsequently propose inference-time software optimizations for layer skipping, and timestep reduction, leveraging Bayesian Optimization with surrogate modeling to perform robust, efficient co-exploration of the joint algorithmic-microarchitectural design spaces under tight computational budgets. Evaluated on both image(ImageNet) and event-based (CIFAR-10 DVS, DVSGesture) classification, the accelerator achieves up to ~467x and ~1.86x energy reduction compared to edge GPU (Jetson Orin Nano) and previous PIM accelerators for spiking transformers, while maintaining competitive task accuracy on ImageNet dataset. This work enables a new class of intelligent ubiquitous edge AI, built using spiking transformer acceleration for low-power, real-time visual processing at the extreme edge.

cs.AR↗

arXiv2025

ScaleDiff: Higher-Resolution Image Synthesis via Efficient and Model-Agnostic Diffusion

Text-to-image diffusion models often exhibit degraded performance when generating images beyond their training resolution. Recent training-free methods can mitigate this limitation, but they often require substantial computation or are incompatible with recent Diffusion Transformer models. In this paper, we propose ScaleDiff, a model-agnostic and highly efficient framework for extending the resolution of pretrained diffusion models without any additional training. A core component of our framework is Neighborhood Patch Attention (NPA), an efficient mechanism that reduces computational redundancy in the self-attention layer with non-overlapping patches. We integrate NPA into an SDEdit pipeline and introduce Latent Frequency Mixing (LFM) to better generate fine details. Furthermore, we apply Structure Guidance to enhance global structure during the denoising process. Experimental results demonstrate that ScaleDiff achieves state-of-the-art performance among training-free methods in terms of both image quality and inference speed on both U-Net and Diffusion Transformer architectures.

cs.LG↗

arXiv2025

Training neural control variates using correlated configurations

Neural control variates (NCVs) have emerged as a powerful tool for variance reduction in Monte Carlo (MC) simulations, particularly in high-dimensional problems where traditional control variates are difficult to construct analytically. By training neural networks to learn auxiliary functions correlated with the target observable, NCVs can significantly reduce estimator variance while preserving unbiasedness. However, a critical but often overlooked aspect of NCV training is the role of autocorrelated samples generated by Markov Chain Monte Carlo (MCMC). While such samples are typically discarded for error estimation due to their statistical redundancy, they may contain useful information about the structure of the underlying probability distribution that can benefit the training process. In this work, we systematically examine the effect of using correlated configurations in training neural control variates. We demonstrate, both conceptually and numerically, that training on correlated data can improve control variate performance, especially in settings with limited computational resources. Our analysis includes empirical results from $U(1)$ gauge theory and scalar field theory, illustrating when and how autocorrelated samples enhance NCV construction. These findings provide practical guidance for the efficient use of MCMC data in training neural networks.

hep-lat↗

arXiv2025

CatchPhrase: EXPrompt-Guided Encoder Adaptation for Audio-to-Image Generation

We propose CatchPhrase, a novel audio-to-image generation framework designed to mitigate semantic misalignment between audio inputs and generated images. While recent advances in multi-modal encoders have enabled progress in cross-modal generation, ambiguity stemming from homographs and auditory illusions continues to hinder accurate alignment. To address this issue, CatchPhrase generates enriched cross-modal semantic prompts (EXPrompt Mining) from weak class labels by leveraging large language models (LLMs) and audio captioning models (ACMs). To address both class-level and instance-level misalignment, we apply multi-modal filtering and retrieval to select the most semantically aligned prompt for each audio sample (EXPrompt Selector). A lightweight mapping network is then trained to adapt pre-trained text-to-image generation models to audio input. Extensive experiments on multiple audio classification datasets demonstrate that CatchPhrase improves audio-to-image alignment and consistently enhances generation quality by mitigating semantic misalignment.

cs.MM↗

arXiv2025

LVLM_CSP: Accelerating Large Vision Language Models via Clustering, Scattering, and Pruning for Reasoning Segmentation

Large Vision Language Models (LVLMs) have been widely adopted to guide vision foundation models in performing reasoning segmentation tasks, achieving impressive performance. However, the substantial computational overhead associated with LVLMs presents a new challenge. The primary source of this computational cost arises from processing hundreds of image tokens. Therefore, an effective strategy to mitigate such overhead is to reduce the number of image tokens, a process known as image token pruning. Previous studies on image token pruning for LVLMs have primarily focused on high level visual understanding tasks, such as visual question answering and image captioning. In contrast, guiding vision foundation models to generate accurate visual masks based on textual queries demands precise semantic and spatial reasoning capabilities. Consequently, pruning methods must carefully control individual image tokens throughout the LVLM reasoning process. Our empirical analysis reveals that existing methods struggle to adequately balance reductions in computational overhead with the necessity to maintain high segmentation accuracy. In this work, we propose LVLM_CSP, a novel training free visual token pruning method specifically designed for LVLM based reasoning segmentation tasks. LVLM_CSP consists of three stages: clustering, scattering, and pruning. Initially, the LVLM performs coarse-grained visual reasoning using a subset of selected image tokens. Next, fine grained reasoning is conducted, and finally, most visual tokens are pruned in the last stage. Extensive experiments demonstrate that LVLM_CSP achieves a 65% reduction in image token inference FLOPs with virtually no accuracy degradation, and a 70% reduction with only a minor 1% drop in accuracy on the 7B LVLM.

cs.CV↗

arXiv2025

Asymptotic errors in adiabatic evolution

The adiabatic theorem in quantum mechanics implies that if a system is in a discrete eigenstate of a Hamiltonian and the Hamiltonian evolves in time arbitrarily slowly, the system will remain in the corresponding eigenstate of the evolved Hamiltonian. Understanding corrections to the adiabatic result that arise when the evolution of the Hamiltonian is slow -- but not arbitrarily slow -- has become increasingly important, especially since adiabatic evolution has been proposed as a method of state preparation in quantum computing. This paper identifies two regimes, an adiabatic regime in which corrections are generically small and can depend on details of the evolution throughout the path, and a hyperadiabatic regime in which the error is given by a form similar to an asymptotic expansion in the inverse of the evolution time with the coefficients depending principally on the behavior at the endpoints. However, the error in this hyperadiabatic regime is neither given by a true asymptotic series nor solely dependent on the endpoints: the coefficients combine the contributions from both endpoints, with relative phase factors that depend on the average spectral gaps along the trajectory, multiplied by the evolution time. The central result of this paper is to identify a quantity, referred to as the typical error, which is obtained by appropriately averaging the error over evolution times that are small compared to the evolution time itself. This typical error is characterized by an asymptotic series and depends solely on the endpoints of the evolution, remaining independent of the details of the intermediate evolution.

quant-ph↗

arXiv2025

VerbDiff: Text-Only Diffusion Models with Enhanced Interaction Awareness

Recent large-scale text-to-image diffusion models generate photorealistic images but often struggle to accurately depict interactions between humans and objects due to their limited ability to differentiate various interaction words. In this work, we propose VerbDiff to address the challenge of capturing nuanced interactions within text-to-image diffusion models. VerbDiff is a novel text-to-image generation model that weakens the bias between interaction words and objects, enhancing the understanding of interactions. Specifically, we disentangle various interaction words from frequency-based anchor words and leverage localized interaction regions from generated images to help the model better capture semantics in distinctive words without extra conditions. Our approach enables the model to accurately understand the intended interaction between humans and objects, producing high-quality images with accurate interactions aligned with specified verbs. Extensive experiments on the HICO-DET dataset demonstrate the effectiveness of our method compared to previous approaches.

cs.GR↗

arXiv2025

Control variates with neural networks

The precision of lattice QCD calculations is often hindered by the stochastic noise inherent in these methods. The control variates method can provide an effective noise reduction but are typically constructed using heuristic approaches, which may be inadequate for complex theories. In this work, we introduce a neural network-based framework for parametrizing control variates, eliminating the reliance on manual guesswork. Using $1+1$ dimensional scalar field theory as a test case, we demonstrate significant variance reduction, particularly in the strong coupling regime. Furthermore, we extend this approach to gauge theories, showcasing its potential to tackle signal-to-noise problems in diverse lattice QCD applications.

hep-lat↗

arXiv2024

Corrections to adiabatic behavior for long paths

The cost and the error of the adiabatic theorem for preparing the final eigenstate are discussed in terms of path length. Previous studies in terms of the norm of the Hamiltonian and its derivatives with the spectral gap are limited in their ability to describe the cost of adiabatic state preparation for certain physically large systems. We argue that total time is not a good measure for determining the computational difficulty of adiabatic quantum computation by developing a no-go theorem. From the result of time-periodic Hamiltonian cases, we suggest that there are proxies for computational cost which typically grow as path length increases when the error is kept fixed and small and consider possible conjectures on how general the behavior is.

quant-ph↗