arXiv subjects

Junyoung Park

Publications and source records attributed to Junyoung Park.

At least 55 records · Page 3Linked to original sources

arXiv2025

VOCABTRIM: Vocabulary Pruning for Efficient Speculative Decoding in LLMs

In this paper, we introduce a simple training-free technique to improve the performance of drafter-based speculative decoding (SpD) methods that incorporates language modeling head (LM head) during drafting process. A drafter-based speculative decoding leverages one or more smaller language models, a.k.a. drafters or draft models, to sample a draft sequence or tree consisting of multiple tokens, followed by verification by a base LLM, a target model, accepting a subset as its valid generation. As it is usually considered that the speculative decoding requires one-to-one mapping between vocabularies of the target model and the draft model, it has been natural to share the vocabulary between them, or even share the LM head as in EAGLE or Medusa. We first identify that this draft token sampling scheme inherently contains an unnecessary inference overhead in drafting, especially for some target LLMs with very large vocabularies. Then, we propose a simple technique, VocabTrim, to mitigate the drafting overhead to improve the generation speed in memory-bound environment. VocabTrim reconstructs the drafter LM head to contain only a limited set of tokens, selected by the most frequently sampled from the vocabulary of the target model. While limiting the vocabulary in drafting slightly degrades the acceptance rate, it significantly reduces the drafting latency in memory-bound process which is often the case on edge devices, resulting in higher memory-bound speed up (MBSU). We show that our method can boost the memory-bound speed-up for Llama-3 models on Spec-Bench, specifically by 16% for Llama-3.2-3B-Instruct.

cs.CL↗

arXiv2025

Generalized Harnack Inequality for Mean Curvature Flow and Ancient Solutions

The goal of this paper is to relax convexity assumption on some classical results in mean curvature flow. In the first half of the paper, we prove a generalized version of Hamilton's differential Harnack inequality which holds for mean convex solutions to mean curvature flow with a lower bound on $\frac{λ_1}{H}$ where $λ_1$ is the smallest principal curvature. Then, we use classical maximum principle to provide several characterizations of family of shrinking spheres for closed, mean convex ancient solution to mean curvature flow with a lower bound on $\frac{λ_1 + .. + λ_k}{H}$ for some $1 \leq k \leq d-1$, where $λ_1 \leq λ_2 \leq .. \leq λ_d$ are the principal curvatures.

math.DG↗

arXiv2024

Expanding Search Space with Diverse Prompting Agents: An Efficient Sampling Approach for LLM Mathematical Reasoning

Large Language Models (LLMs) have exhibited remarkable capabilities in many complex tasks including mathematical reasoning. However, traditional approaches heavily rely on ensuring self-consistency within single prompting method, which limits the exploration of diverse problem-solving strategies. This study addresses these limitations by performing an experimental analysis of distinct prompting methods within the domain of mathematical reasoning. Our findings demonstrate that each method explores a distinct search space, and this differentiation becomes more evident with increasing problem complexity. To leverage this phenomenon, we applied efficient sampling process that uniformly combines samples from these diverse methods, which not only expands the maximum search space but achieves higher performance with fewer runs compared to single methods. Especially, within the subset of difficult questions of MATH dataset named MATH-hard, The maximum search space was achieved while utilizing approximately 43% fewer runs than single methods on average. These findings highlight the importance of integrating diverse problem-solving strategies to enhance the reasoning abilities of LLMs.

cs.CL↗

arXiv2024

Towards Efficient Visual-Language Alignment of the Q-Former for Visual Reasoning Tasks

Recent advancements in large language models have demonstrated enhanced capabilities in visual reasoning tasks by employing additional encoders for aligning different modalities. While the Q-Former has been widely used as a general encoder for aligning several modalities including image, video, audio, and 3D with large language models, previous works on its efficient training and the analysis of its individual components have been limited. In this work, we investigate the effectiveness of parameter efficient fine-tuning (PEFT) the Q-Former using InstructBLIP with visual reasoning benchmarks ScienceQA and IconQA. We observe that applying PEFT to the Q-Former achieves comparable performance to full fine-tuning using under 2% of the trainable parameters. Additionally, we employ AdaLoRA for dynamic parameter budget reallocation to examine the relative importance of the Q-Former's sublayers with 4 different benchmarks. Our findings reveal that the self-attention layers are noticeably more important in perceptual visual-language reasoning tasks, and relative importance of FFN layers depends on the complexity of visual-language patterns involved in tasks. The code is available at https://github.com/AttentionX/InstructBLIP_PEFT.

cs.CL↗

arXiv2024

Topological Floquet engineering of a three-band optical lattice with dual-mode resonant driving

We present a Floquet framework for controlling topological features of a one-dimensional optical lattice system with dual-mode resonant driving, in which both the amplitude and phase of the lattice potential are modulated simultaneously. We investigate a three-band model consisting of the three lowest orbitals and elucidate the formation of a cross-linked two-leg ladder through an indirect interband coupling via an off-resonant band. We numerically demonstrate the emergence of topologically nontrivial bands within the driven system, and a topological charge pumping phenomenon with cyclic parameter changes in the dual-mode resonant driving. Finally, we show that the band topology in the driven three-band system is protected by parity-time reversal symmetry.

cond-mat.quant-gas↗

arXiv2024

Genetic Algorithms with Neural Cost Predictor for Solving Hierarchical Vehicle Routing Problems

When vehicle routing decisions are intertwined with higher-level decisions, the resulting optimization problems pose significant challenges for computation. Examples are the multi-depot vehicle routing problem (MDVRP), where customers are assigned to depots before delivery, and the capacitated location routing problem (CLRP), where the locations of depots should be determined first. A simple and straightforward approach for such hierarchical problems would be to separate the higher-level decisions from the complicated vehicle routing decisions. For each higher-level decision candidate, we may evaluate the underlying vehicle routing problems to assess the candidate. As this approach requires solving vehicle routing problems multiple times, it has been regarded as impractical in most cases. We propose a novel deep-learning-based approach called Genetic Algorithm with Neural Cost Predictor (GANCP) to tackle the challenge and simplify algorithm developments. For each higher-level decision candidate, we predict the objective function values of the underlying vehicle routing problems using a pre-trained graph neural network without actually solving the routing problems. In particular, our proposed neural network learns the objective values of the HGS-CVRP open-source package that solves capacitated vehicle routing problems. Our numerical experiments show that this simplified approach is effective and efficient in generating high-quality solutions for both MDVRP and CLRP and has the potential to expedite algorithm developments for complicated hierarchical problems. We provide computational results evaluated in the standard benchmark instances used in the literature.

cs.NE↗

arXiv2024

Multi-stream deep learning framework to predict mild cognitive impairment with Rey Complex Figure Test

Drawing tests like the Rey Complex Figure Test (RCFT) are widely used to assess cognitive functions such as visuospatial skills and memory, making them valuable tools for detecting mild cognitive impairment (MCI). Despite their utility, existing predictive models based on these tests often suffer from limitations like small sample sizes and lack of external validation, which undermine their reliability. We developed a multi-stream deep learning framework that integrates two distinct processing streams: a multi-head self-attention based spatial stream using raw RCFT images and a scoring stream employing a previously developed automated scoring system. Our model was trained on data from 1,740 subjects in the Korean cohort and validated on an external hospital dataset of 222 subjects from Korea. The proposed multi-stream model demonstrated superior performance over baseline models (AUC = 0.872, Accuracy = 0.781) in external validation. The integration of both spatial and scoring streams enables the model to capture intricate visual details from the raw images while also incorporating structured scoring data, which together enhance its ability to detect subtle cognitive impairments. This dual approach not only improves predictive accuracy but also increases the robustness of the model, making it more reliable in diverse clinical settings. Our model has practical implications for clinical settings, where it could serve as a cost-effective tool for early MCI screening.

cs.CV↗

arXiv2024

Token-Picker: Accelerating Attention in Text Generation with Minimized Memory Transfer via Probability Estimation

The attention mechanism in text generation is memory-bounded due to its sequential characteristics. Therefore, off-chip memory accesses should be minimized for faster execution. Although previous methods addressed this by pruning unimportant tokens, they fall short in selectively removing tokens with near-zero attention probabilities in each instance. Our method estimates the probability before the softmax function, effectively removing low probability tokens and achieving an 12.1x pruning ratio without fine-tuning. Additionally, we present a hardware design supporting seamless on-demand off-chip access. Our approach shows 2.6x reduced memory accesses, leading to an average 2.3x speedup and a 2.4x energy efficiency.

cs.AR↗

arXiv2024

Enhancing Source-Free Domain Adaptive Object Detection with Low-confidence Pseudo Label Distillation

Source-Free domain adaptive Object Detection (SFOD) is a promising strategy for deploying trained detectors to new, unlabeled domains without accessing source data, addressing significant concerns around data privacy and efficiency. Most SFOD methods leverage a Mean-Teacher (MT) self-training paradigm relying heavily on High-confidence Pseudo Labels (HPL). However, these HPL often overlook small instances that undergo significant appearance changes with domain shifts. Additionally, HPL ignore instances with low confidence due to the scarcity of training samples, resulting in biased adaptation toward familiar instances from the source domain. To address this limitation, we introduce the Low-confidence Pseudo Label Distillation (LPLD) loss within the Mean-Teacher based SFOD framework. This novel approach is designed to leverage the proposals from Region Proposal Network (RPN), which potentially encompasses hard-to-detect objects in unfamiliar domains. Initially, we extract HPL using a standard pseudo-labeling technique and mine a set of Low-confidence Pseudo Labels (LPL) from proposals generated by RPN, leaving those that do not overlap significantly with HPL. These LPL are further refined by leveraging class-relation information and reducing the effect of inherent noise for the LPLD loss calculation. Furthermore, we use feature distance to adaptively weight the LPLD loss to focus on LPL containing a larger foreground area. Our method outperforms previous SFOD methods on four cross-domain object detection benchmarks. Extensive experiments demonstrate that our LPLD loss leads to effective adaptation by reducing false negatives and facilitating the use of domain-invariant knowledge from the source model. Code is available at https://github.com/junia3/LPLD.

cs.CV↗

arXiv2024

Direct Alignment of Draft Model for Speculative Decoding with Chat-Fine-Tuned LLMs

Text generation with Large Language Models (LLMs) is known to be memory bound due to the combination of their auto-regressive nature, huge parameter counts, and limited memory bandwidths, often resulting in low token rates. Speculative decoding has been proposed as a solution for LLM inference acceleration. However, since draft models are often unavailable in the modern open-source LLM families, e.g., for Llama 2 7B, training a high-quality draft model is required to enable inference acceleration via speculative decoding. In this paper, we propose a simple draft model training framework for direct alignment to chat-capable target models. With the proposed framework, we train Llama 2 Chat Drafter 115M, a draft model for Llama 2 Chat 7B or larger, with only 1.64\% of the original size. Our training framework only consists of pretraining, distillation dataset generation, and finetuning with knowledge distillation, with no additional alignment procedure. For the finetuning step, we use instruction-response pairs generated by target model for distillation in plausible data distribution, and propose a new Total Variation Distance++ (TVD++) loss that incorporates variance reduction techniques inspired from the policy gradient method in reinforcement learning. Our empirical results show that Llama 2 Chat Drafter 115M with speculative decoding achieves up to 2.3 block efficiency and 2.4$\times$ speed-up relative to autoregressive decoding on various tasks with no further task-specific fine-tuning.

cs.LG↗

arXiv2024

On Speculative Decoding for Multimodal Large Language Models

Inference with Multimodal Large Language Models (MLLMs) is slow due to their large-language-model backbone which suffers from memory bandwidth bottleneck and generates tokens auto-regressively. In this paper, we explore the application of speculative decoding to enhance the inference efficiency of MLLMs, specifically the LLaVA 7B model. We show that a language-only model can serve as a good draft model for speculative decoding with LLaVA 7B, bypassing the need for image tokens and their associated processing components from the draft model. Our experiments across three different tasks show that speculative decoding can achieve a memory-bound speedup of up to 2.37$\times$ using a 115M parameter language model that we trained from scratch. Additionally, we introduce a compact LLaVA draft model incorporating an image adapter, which shows marginal performance gains in image captioning while maintaining comparable results in other tasks.

cs.CL↗

arXiv2024

Recursive Speculative Decoding: Accelerating LLM Inference via Sampling Without Replacement

Speculative decoding is an inference-acceleration method for large language models (LLMs) where a small language model generates a draft-token sequence which is further verified by the target LLM in parallel. Recent works have advanced this method by establishing a draft-token tree, achieving superior performance over a single-sequence speculative decoding. However, those works independently generate tokens at each level of the tree, not leveraging the tree's entire diversifiability. Besides, their empirical superiority has been shown for fixed length of sequences, implicitly granting more computational resource to LLM for the tree-based methods. None of the existing works has conducted empirical studies with fixed target computational budgets despite its importance to resource-bounded devices. We present Recursive Speculative Decoding (RSD), a novel tree-based method that samples draft tokens without replacement and maximizes the diversity of the tree. During RSD's drafting, the tree is built by either Gumbel-Top-$k$ trick that draws tokens without replacement in parallel or Stochastic Beam Search that samples sequences without replacement while early-truncating unlikely draft sequences and reducing the computational cost of LLM. We empirically evaluate RSD with Llama 2 and OPT models, showing that RSD outperforms the baseline methods, consistently for fixed draft sequence length and in most cases for fixed computational budgets at LLM.

cs.LG↗

arXiv2023

Layer-wise Auto-Weighting for Non-Stationary Test-Time Adaptation

Given the inevitability of domain shifts during inference in real-world applications, test-time adaptation (TTA) is essential for model adaptation after deployment. However, the real-world scenario of continuously changing target distributions presents challenges including catastrophic forgetting and error accumulation. Existing TTA methods for non-stationary domain shifts, while effective, incur excessive computational load, making them impractical for on-device settings. In this paper, we introduce a layer-wise auto-weighting algorithm for continual and gradual TTA that autonomously identifies layers for preservation or concentrated adaptation. By leveraging the Fisher Information Matrix (FIM), we first design the learning weight to selectively focus on layers associated with log-likelihood changes while preserving unrelated ones. Then, we further propose an exponential min-max scaler to make certain layers nearly frozen while mitigating outliers. This minimizes forgetting and error accumulation, leading to efficient adaptation to non-stationary target distribution. Experiments on CIFAR-10C, CIFAR-100C, and ImageNet-C show our method outperforms conventional continual and gradual TTA approaches while significantly reducing computational load, highlighting the importance of FIM-based learning weight in adapting to continuously or gradually shifting target domains.

cs.LG↗

arXiv2023

Generating Dispatching Rules for the Interrupting Swap-Allowed Blocking Job Shop Problem Using Graph Neural Network and Reinforcement Learning

The interrupting swap-allowed blocking job shop problem (ISBJSSP) is a complex scheduling problem that is able to model many manufacturing planning and logistics applications realistically by addressing both the lack of storage capacity and unforeseen production interruptions. Subjected to random disruptions due to machine malfunction or maintenance, industry production settings often choose to adopt dispatching rules to enable adaptive, real-time re-scheduling, rather than traditional methods that require costly re-computation on the new configuration every time the problem condition changes dynamically. To generate dispatching rules for the ISBJSSP problem, we introduce a dynamic disjunctive graph formulation characterized by nodes and edges subjected to continuous deletions and additions. This formulation enables the training of an adaptive scheduler utilizing graph neural networks and reinforcement learning. Furthermore, a simulator is developed to simulate interruption, swapping, and blocking in the ISBJSSP setting. Employing a set of reported benchmark instances, we conduct a detailed experimental study on ISBJSSP instances with a range of machine shutdown probabilities to show that the scheduling policies generated can outperform or are at least as competitive as existing dispatching rules with predetermined priority. This study shows that the ISBJSSP, which requires real-time adaptive solutions, can be scheduled efficiently with the proposed method when production interruptions occur with random machine shutdowns.

cs.LG↗

arXiv2023

Sym-NCO: Leveraging Symmetricity for Neural Combinatorial Optimization

Deep reinforcement learning (DRL)-based combinatorial optimization (CO) methods (i.e., DRL-NCO) have shown significant merit over the conventional CO solvers as DRL-NCO is capable of learning CO solvers less relying on problem-specific expert domain knowledge (heuristic method) and supervised labeled data (supervised learning method). This paper presents a novel training scheme, Sym-NCO, which is a regularizer-based training scheme that leverages universal symmetricities in various CO problems and solutions. Leveraging symmetricities such as rotational and reflectional invariance can greatly improve the generalization capability of DRL-NCO because it allows the learned solver to exploit the commonly shared symmetricities in the same CO problem class. Our experimental results verify that our Sym-NCO greatly improves the performance of DRL-NCO methods in four CO tasks, including the traveling salesman problem (TSP), capacitated vehicle routing problem (CVRP), prize collecting TSP (PCTSP), and orienteering problem (OP), without utilizing problem-specific expert domain knowledge. Remarkably, Sym-NCO outperformed not only the existing DRL-NCO methods but also a competitive conventional solver, the iterative local search (ILS), in PCTSP at 240 faster speed. Our source code is available at https://github.com/alstn12088/Sym-NCO.

cs.LG↗

arXiv2022

Learning context-aware adaptive solvers to accelerate quadratic programming

Convex quadratic programming (QP) is an important sub-field of mathematical optimization. The alternating direction method of multipliers (ADMM) is a successful method to solve QP. Even though ADMM shows promising results in solving various types of QP, its convergence speed is known to be highly dependent on the step-size parameter $ρ$. Due to the absence of a general rule for setting $ρ$, it is often tuned manually or heuristically. In this paper, we propose CA-ADMM (Context-aware Adaptive ADMM)) which learns to adaptively adjust $ρ$ to accelerate ADMM. CA-ADMM extracts the spatio-temporal context, which captures the dependency of the primal and dual variables of QP and their temporal evolution during the ADMM iterations. CA-ADMM chooses $ρ$ based on the extracted context. Through extensive numerical experiments, we validated that CA-ADMM effectively generalizes to unseen QP problems with different sizes and classes (i.e., having different QP parameter structures). Furthermore, we verified that CA-ADMM could dynamically adjust $ρ$ considering the stage of the optimization process to accelerate the convergence speed further.

math.OC↗

arXiv2022

Development of IoT Sensor and Cloud-based Server for Cloud-based Bridge Long-term Monitoring

As social infrastructures rapidly age, it is crucial to create a digital SOC (Social Overhead Capital) maintenance system for preventive maintenance. Using IoT sensors installed on the structures, abnormal signals produced by the structures may be rapidly identified and the best judgments can be made. In this study, a multimetric IoT sensor was integrated into a digital SOC monitoring system for long-term monitoring, use in a cloud computing server for automated and powerful data analysis, and for creating databases to perform: (1) multimetric sensing, (2) long-term operation, and (3) LTE-based direct communication. The created sensor included five strain sensing axes and three acceleration axes, and it also had an event-driven power management system that only turned on the sensors when the vibration reached a certain threshold or the timer started. Long-term operation might be made possible by the power management system and an extra solar panel charging. Real-time data transmission to the server from the sensors was accomplished using low-power LTE CAT.1 connection, which does not need an additional gateway device. Additionally, a displacement fusion technique was created on the cloud server to get reference-free structural displacement for ambient structural assessment after receiving multi-variable data from the sensor. A steel railroad and concrete girder bridge were used to experimentally evaluate the suggested digital SOC system.

cs.HC↗

arXiv2022

Development of AI-cloud based high-sensitivity wireless smart sensor for port structure monitoring

Regular structural monitoring of port structure is crucial to cope with rapid degeneration owing to its exposure to saline and collisional environment. However, most of the inspections are being done visually by human in irregular-basis. To overcome the complication, lots of research related to vibration-based monitoring system with sensor has been devised. Nonetheless, it was difficult to measure ambient vibration due to port's diminutive amplitude and specify the exact timing of berthing, which is the major excitation source. This study developed a novel cloud-AI based wireless sensor system with high-sensitivity accelerometer M-A352, which has 0.2uG/sqrt(Hz) noise density, 0.003mg of ultra-low noise feature, and 1000Hz of sampling frequency. The sensor is triggered based on either predefined schedule or long rangefinder. After that, the detection of ship is done by AI object detection technique called Faster R-CNN with backbone network of ResNet for the convolution part. Coordinate and size of the detected anchor box is further processed to certify the berthing ship. Collected data are automatically sent to the cloud server through LTE CAT 1 modem within 10Mbps. The system was installed in the actual port field in Korea for few days as a preliminary investigation of proposed system. Additionally, acceleration, slope, and temperature data are analyzed to suggest the possibility of vibration-based port condition assessment.

eess.SP↗