arXiv subjects

Hieu Nguyen

Publications and source records attributed to Hieu Nguyen.

At least 19 recordsLinked to original sources

arXiv2026

Token-Level Diagnosis of Sycophancy in LLMs with Attribution-Guided Steering

Sycophancy refers to the tendency for large language models (LLMs) to match user beliefs at the cost of factual correctness, thereby undermining model reliability. Prior work on evaluating sycophancy in LLMs aims to assess whether a model's output matches an authority's claim, but cannot reveal which part of the prompt drives this sycophantic behavior. To bridge this gap, we investigate the relationship of sycophantic responses with an authority's credentials, their assertive claim, and the problem statement. We introduce the Authority Share Index (ASI), an Integrated Gradients-based token attribution method, which measures the degree to which a model's decision is driven by authority-related text. Through extensive experiments across five models and 30 test configurations, we find that sycophantic responses consistently direct more attention toward authority tokens than resistant ones. Moreover, our token attribution method reveals that for the sycophantic cases, the claim asserted by the authority receives more attention than the authority's credentials. Building on these findings, we propose attribution-guided contrastive activation steering to mitigate LLM sycophancy. Our method constructs a steering vector from high-attribution tokens of sycophantic and resistant responses, selectively pushing models toward resistance. This enables inference-time steering without retraining, lowering sycophancy from 96% to 25% in the strongest case. Together, our results show that token-level attribution can both explain what drives sycophancy and directly inform a practical intervention.

cs.CL↗

arXiv2026

DiStash: A Disaggregated Multi-Stash Transactional Key-Value Store

A stash is a storage medium such as Dynamic Random Access Memory (DRAM), Solid State Disk (SSD), Hard Disk Drive (HDD), or Non-Volatile Memory (NVM). This paper presents a disaggregated transactional key-value (KV) store, DiStash, that governs KVs cross pools of stash types. It enables an application to use a single transaction to read and write different copies of one or more key-value pair across the different pools of stashes. It simplifies the application logic by (a) preventing undesirable race conditions that may cause copies of data across different stash pools to reflect different values and/or (b) failures that may result in loss of key-value pairs. A configuration of DiStash may use a pool of stashes as either ephemeral or durable storage. The application dictates whether the content of its participating stashes are inclusive (replicated) or exclusive (tiered). We implement a DiStash by extending FoundationDB. We quantify the tradeoffs with its design decisions using microbenchmarks and eBay's production workload. We open source our implementation at https://github.com/ebay-USC/DiStash.

cs.DB↗

arXiv2026

VisChronos: Revolutionizing Image Captioning Through Real-Life Events

This paper aims to bridge the semantic gap between visual content and natural language understanding by leveraging historical events in the real world as a source of knowledge for caption generation. We propose VisChronos, a novel framework that utilizes large language models and dense captioning models to identify and describe real-life events from a single input image. Our framework can automatically generate detailed and context-aware event descriptions, enhancing the descriptive quality and contextual relevance of generated captions to address the limitations of traditional methods in capturing contextual narratives. Furthermore, we introduce a new dataset, EventCap (https://zenodo.org/records/14004909), specifically constructed using the proposed framework, designed to enhance the model's ability to identify and understand complex events. The user study demonstrates the efficacy of our solution in generating accurate, coherent, and event-focused descriptions, paving the way for future research in event-centric image understanding.

cs.CV↗

arXiv2026

ByteRover: Agent-Native Memory Through LLM-Curated Hierarchical Context

Memory-Augmented Generation (MAG) extends large language models with external memory to support long-context reasoning, but existing approaches universally treat memory as an external service that agents call into, delegating storage to separate pipelines of chunking, embedding, and graph extraction. This architectural separation means the system that stores knowledge does not understand it, leading to semantic drift between what the agent intended to remember and what the pipeline actually captured, loss of coordination context across agents, and fragile recovery after failures. In this paper, we propose ByteRover, an agent-native memory architecture that inverts the memory pipeline: the same LLM that reasons about a task also curates, structures, and retrieves knowledge. ByteRover represents knowledge in a hierarchical Context Tree, a file-based knowledge graph organized as Domain, Topic, Subtopic, and Entry, where each entry carries explicit relations, provenance, and an Adaptive Knowledge Lifecycle (AKL) with importance scoring, maturity tiers, and recency decay. Retrieval uses a 5-tier progressive strategy that resolves most queries at sub-100 ms latency without LLM calls, escalating to agentic reasoning only for novel questions. Experiments on LoCoMo and LongMemEval demonstrate that ByteRover achieves state-of-the-art accuracy on LoCoMo and competitive results on LongMemEval while requiring zero external infrastructure, no vector database, no graph database, no embedding service, with all knowledge stored as human-readable markdown files on the local filesystem.

cs.AI↗

arXiv2025

Wine Characterisation with Spectral Information and Predictive Artificial Intelligence

The purpose of this paper is to use absorbance data obtained by human tasting and an ultraviolet-visible (UV-Vis) scanning spectrophotometer to predict the attributes of grape juice (GJ) and to classify the wine's origin, respectively. The approach combined machine learning (ML) techniques with spectroscopy to find a relatively simple way to apply them in two stages of winemaking and help improve the traditional wine analysis methods regarding sensory data and wine's origins. This new technique has overcome the disadvantages of the complex sensors by taking advantage of spectral fingerprinting technology and forming a comprehensive study of the employment of AI in the wine analysis domain. In the results, Support Vector Machine (SVM) was the most efficient and robust in both attributes and origin prediction tasks. Both the accuracy and F1 score of the origin prediction exceed 91%. The feature ranking approach found that the more influential wavelengths usually appear at the lower end of the scan range, 250 nm (nanometers) to 420 nm, which is believed to be of great help for selecting appropriate validation methods and sensors to extract wine data in future research. The knowledge of this research provides new ideas and early solutions for the wine industry or other beverage industries to integrate big data and IoT in the future, which significantly promotes the development of 'Smart Wineries'.

cs.LG↗

arXiv2025

OpenEvents V1: Large-Scale Benchmark Dataset for Multimodal Event Grounding

We introduce OpenEvents V1a large-scale benchmark dataset designed to advance event-centric vision-language understanding. Unlike conventional image captioning and retrieval datasets that focus on surface-level descriptions, OpenEvents V1 dataset emphasizes contextual and temporal grounding through three primary tasks: (1) generating rich, event-aware image captions, (2) retrieving event-relevant news articles from image queries, and (3) retrieving event-relevant images from narrative-style textual queries. The dataset comprises over 200,000 news articles and 400,000 associated images sourced from CNN and The Guardian, spanning diverse domains and time periods. We provide extensive baseline results and standardized evaluation protocols for all tasks. OpenEvents V1 establishes a robust foundation for developing multimodal AI systems capable of deep reasoning over complex real-world events. The dataset is publicly available at https://ltnghia.github.io/eventa/openevents-v1.

cs.CV↗

arXiv2025

BSA: Ball Sparse Attention for Large-scale Geometries

Self-attention scales quadratically with input size, limiting its use for large-scale physical systems. Although sparse attention mechanisms provide a viable alternative, they are primarily designed for regular structures such as text or images, making them inapplicable for irregular geometries. In this work, we present Ball Sparse Attention (BSA), which adapts Native Sparse Attention (NSA) (Yuan et al., 2025) to unordered point sets by imposing regularity using the Ball Tree structure from the Erwin Transformer (Zhdanov et al., 2025). We modify NSA's components to work with ball-based neighborhoods, yielding a global receptive field at sub-quadratic cost. On an airflow pressure prediction task, we achieve accuracy comparable to Full Attention while significantly reducing the theoretical computational complexity. Our implementation is available at https://github.com/britacatalin/bsa.

cs.LG↗

arXiv2025

Smoothing Out Hallucinations: Mitigating LLM Hallucination with Smoothed Knowledge Distillation

Large language models (LLMs) often suffer from hallucination, generating factually incorrect or ungrounded content, which limits their reliability in high-stakes applications. A key factor contributing to hallucination is the use of hard labels during training, which enforce deterministic supervision, encourage overconfidence, and disregard the uncertainty inherent in natural language. To address this, we propose mitigating hallucination through knowledge distillation (KD), where a teacher model provides smoothed soft labels to a student model, reducing overconfidence and improving factual grounding. We apply KD during supervised finetuning on instructional data, evaluating its effectiveness across LLMs from different families. Experimental results on summarization benchmarks demonstrate that KD reduces hallucination compared to standard finetuning while preserving performance on general NLP tasks. These findings highlight KD as a promising approach for mitigating hallucination in LLMs and improving model reliability.

cs.CL↗

arXiv2024

One-Hop Sub-Query Result Caches for Graph Database Systems

This paper introduces a novel one-hop sub-query result cache for processing graph read transactions, gR-Txs, in a graph database system. The one-hop navigation is from a vertex using either its in-coming or out-going edges with selection predicates that filter edges and vertices. Its cache entry identifies a unique one-hop sub-query (key) and its result set consisting of immutable vertex ids (value). When processing a gR-Tx, the query processor identifies its sequence of individual one-hop sub-queries and looks up their results in the cache. A cache hit fetches less data from the storage manager and eliminates the requirement to process the one-hop sub-query. A cache miss populates the cache asynchronously and in a transactional manner, maintaining the separation of read and write paths of our transactional storage manager. A graph read and write transaction, gRW-Tx, identifies the impacted cache entries and either deletes or updates them. Our implementation of the cache is inside the graph query processing engine and transparent to a user application. We evaluate the cache using our eCommerce production workload and with rules that re-write graph queries to maximize the performance enhancements observed with the cache. Obtained results show the cache enhances 95th and 99th percentile of query response times by at least 2x and 1.63x, respectively. When combined with query re-writing, the enhancements are at least 2.33x and 4.48x, respectively. An interesting result is the significant performance enhancement observed by the indirect beneficiaries of the cache, gRW-Txs and gR-Txs that do not reference one-hop sub-queries. The cache frees system resources to expedite their processing significantly.

cs.DB↗

arXiv2024

"A Woman is More Culturally Knowledgeable than A Man?": The Effect of Personas on Cultural Norm Interpretation in LLMs

As the deployment of large language models (LLMs) expands, there is an increasing demand for personalized LLMs. One method to personalize and guide the outputs of these models is by assigning a persona -- a role that describes the expected behavior of the LLM (e.g., a man, a woman, an engineer). This study investigates whether an LLM's understanding of social norms varies across assigned personas. Ideally, the perception of a social norm should remain consistent regardless of the persona, since acceptability of a social norm should be determined by the region the norm originates from, rather than by individual characteristics such as gender, body size, or race. A norm is universal within its cultural context. In our research, we tested 36 distinct personas from 12 sociodemographic categories (e.g., age, gender, beauty) across four different LLMs. We find that LLMs' cultural norm interpretation varies based on the persona used and the norm interpretation also varies within a sociodemographic category (e.g., a fat person and a thin person as in physical appearance group) where an LLM with the more socially desirable persona (e.g., a thin person) interprets social norms more accurately than with the less socially desirable persona (e.g., a fat person). We also discuss how different types of social biases may contribute to the results that we observe.

cs.CL↗

arXiv2024

Ensemble Learning for Vietnamese Scene Text Spotting in Urban Environments

This paper presents a simple yet efficient ensemble learning framework for Vietnamese scene text spotting. Leveraging the power of ensemble learning, which combines multiple models to yield more accurate predictions, our approach aims to significantly enhance the performance of scene text spotting in challenging urban settings. Through experimental evaluations on the VinText dataset, our proposed method achieves a significant improvement in accuracy compared to existing methods with an impressive accuracy of 5%. These results unequivocally demonstrate the efficacy of ensemble learning in the context of Vietnamese scene text spotting in urban environments, highlighting its potential for real world applications, such as text detection and recognition in urban signage, advertisements, and various text-rich urban scenes.

cs.CV↗

arXiv2024

Physics-based material parameters extraction from perovskite experiments via Bayesian optimization

The ability to extract material parameters of perovskite from quantitative experimental analysis is essential for rational design of photovoltaic and optoelectronic applications. However, the difficulty of this analysis increases significantly with the complexity of the theoretical model and the number of material parameters for perovskite. Here we use Bayesian optimization to develop an analysis platform that can extract up to 8 fundamental material parameters of an organometallic perovskite semiconductor from a transient photoluminescence experiment, based on a complex full physics model that includes drift-diffusion of carriers and dynamic defect occupation. An example study of thermal degradation reveals that the carrier mobility and trap-assisted recombination coefficient are reduced noticeably, while the defect energy level remains nearly unchanged. The reduced carrier mobility can dominate the overall effect on thermal degradation of perovskite solar cells by reducing the fill factor, despite the opposite effect of the reduced trap-assisted recombination coefficient on increasing the fill factor. In future, this platform can be conveniently applied to other experiments or to combinations of experiments, accelerating materials discovery and optimization of semiconductor materials for photovoltaics and other applications.

cond-mat.mtrl-sci↗

arXiv2024

Personalizing Driver Safety Interfaces via Driver Cognitive Factors Inference

Recent advances in AI and intelligent vehicle technology hold promise to revolutionize mobility and transportation, in the form of advanced driving assistance (ADAS) interfaces. Although it is widely recognized that certain cognitive factors, such as impulsivity and inhibitory control, are related to risky driving behavior, play a significant role in on-road risk-taking, existing systems fail to leverage such factors. Varying levels of these cognitive factors could influence the effectiveness and acceptance of driver safety interfaces. We demonstrate an approach for personalizing driver interaction via driver safety interfaces that are triggered based on a learned recurrent neural network. The network is trained from a population of human drivers to infer impulsivity and inhibitory control from recent driving behavior. Using a high-fidelity vehicle motion simulator, we demonstrate the ability to deduce these factors from driver behavior. We then use these inferred factors to make instantaneous determinations on whether or not to engage a driver safety interface. This interface aims to decrease a driver's speed during yellow lights and reduce their inclination to run through them.

cs.HC↗

arXiv2023

Real-Time Traffic Sign Detection: A Case Study in a Santa Clara Suburban Neighborhood

This research project aims to develop a real-time traffic sign detection system using the YOLOv5 architecture and deploy it for efficient traffic sign recognition during a drive in a suburban neighborhood. The project's primary objectives are to train the YOLOv5 model on a diverse dataset of traffic sign images and deploy the model on a suitable hardware platform capable of real-time inference. The project will involve collecting a comprehensive dataset of traffic sign images. By leveraging the trained YOLOv5 model, the system will detect and classify traffic signs from a real-time camera on a dashboard inside a vehicle. The performance of the deployed system will be evaluated based on its accuracy in detecting traffic signs, real-time processing speed, and overall reliability. During a case study in a suburban neighborhood, the system demonstrated a notable 96% accuracy in detecting traffic signs. This research's findings have the potential to improve road safety and traffic management by providing timely and accurate real-time information about traffic signs and can pave the way for further research into autonomous driving.

cs.CV↗

arXiv2023

ESWORD: Implementation of Wireless Jamming Attacks in a Real-World Emulated Network

Wireless jamming attacks have plagued wireless communication systems and will continue to do so going forward with technological advances. These attacks fall under the category of Electronic Warfare (EW), a continuously growing area in both attack and defense of the electromagnetic spectrum, with one subcategory being electronic attacks. Jamming attacks fall under this specific subcategory of EW as they comprise adversarial signals that attempt to disrupt, deny, degrade, destroy, or deceive legitimate signals in the electromagnetic spectrum. While jamming is not going away, recent research advances have started to get the upper hand against these attacks by leveraging new methods and techniques, such as machine learning. However, testing such jamming solutions on a wide and realistic scale is a daunting task due to strict regulations on spectrum emissions. In this paper, we introduce eSWORD, the first large-scale framework that allows users to safely conduct real-time and controlled jamming experiments with hardware-in-the-loop. This is done by integrating eSWORD into the Colosseum wireless network emulator that enables large-scale experiments with up to 50 software-defined radio nodes. We compare the performance of eSWORD with that of real-world jamming systems by using an over-the-air wireless testbed (ensuring safe measures were taken when conducting experiments). Our experimental results demonstrate that eSWORD follows similar patterns in throughput, signal-to-noise ratio, and link status to real-world jamming experiments, testifying to the high accuracy of the emulated eSWORD setup.

cs.NI↗

arXiv2022

Deep Variational Inverse Scattering

Inverse medium scattering solvers generally reconstruct a single solution without an associated measure of uncertainty. This is true both for the classical iterative solvers and for the emerging deep learning methods. But ill-posedness and noise can make this single estimate inaccurate or misleading. While deep networks such as conditional normalizing flows can be used to sample posteriors in inverse problems, they often yield low-quality samples and uncertainty estimates. In this paper, we propose U-Flow, a Bayesian U-Net based on conditional normalizing flows, which generates high-quality posterior samples and estimates physically-meaningful uncertainty. We show that the proposed model significantly outperforms the recent normalizing flows in terms of posterior sample quality while having comparable performance with the U-Net in point estimation.

cs.LG↗

arXiv2022

Predicting housing prices and analyzing real estate market in the Chicago suburbs using Machine Learning

The pricing of housing properties is determined by a variety of factors. However, post-pandemic markets have experienced volatility in the Chicago suburb area, which have affected house prices greatly. In this study, analysis was done on the Naperville/Bolingbrook real estate market to predict property prices based on these housing attributes through machine learning models, and to evaluate the effectiveness of such models in a volatile market space. Gathering data from Redfin, a real estate website, sales data from 2018 up until the summer season of 2022 were collected for research. By analyzing these sales in this range of time, we can also look at the state of the housing market and identify trends in price. For modeling the data, the models used were linear regression, support vector regression, decision tree regression, random forest regression, and XGBoost regression. To analyze results, comparison was made on the MAE, RMSE, and R-squared values for each model. It was found that the XGBoost model performs the best in predicting house prices despite the additional volatility sponsored by post-pandemic conditions. After modeling, Shapley Values (SHAP) were used to evaluate the weights of the variables in constructing models.

cs.LG↗

arXiv2022

MTet: Multi-domain Translation for English and Vietnamese

We introduce MTet, the largest publicly available parallel corpus for English-Vietnamese translation. MTet consists of 4.2M high-quality training sentence pairs and a multi-domain test set refined by the Vietnamese research community. Combining with previous works on English-Vietnamese translation, we grow the existing parallel dataset to 6.2M sentence pairs. We also release the first pretrained model EnViT5 for English and Vietnamese languages. Combining both resources, our model significantly outperforms previous state-of-the-art results by up to 2 points in translation BLEU score, while being 1.6 times smaller.

cs.CL↗