arXiv subjects

Thanh-Son Nguyen

Publications and source records attributed to Thanh-Son Nguyen.

11 recordsLinked to original sources

arXiv2026

GGC: Selective Query Correction for Reliable Text-to-SPARQL Generation

Large language models (LLMs) have demonstrated strong capabilities in structured query generation, making them a natural choice for Text-to-SPARQL, which translates natural language questions into executable SPARQL queries over knowledge graphs. However, their initial outputs remain unreliable: generated queries may be executable yet semantically misaligned with input questions, leading to incorrect retrieval. To address this issue, we propose Generator-Gate-Corrector (GGC), a framework for reliable LLM-based Text-to-SPARQL generation. GGC first uses a Generator to produce an initial query, then applies a Gate to predict whether correction is needed, and finally invokes a Corrector only for selected high-risk queries. This selective correction mechanism avoids unnecessary modifications and reduces the risk of degrading originally correct queries. Experiments on MCQA show that GGC improves query-level accuracy from 90.23\% to 98.33\% while reducing inference overhead by 45\% compared with correcting all generated queries. Ablation studies show that the Gate is robust across thresholds and that Corrector training data composition affects correction effectiveness and stability. Overall, the results demonstrate that selective correction enhances the accuracy, reliability, and efficiency of LLM-based text-to-SPARQL generation.

cs.CL↗

arXiv2026

Analytical Retrieval of Material Parameters in Monolayer Transition-Metal Dichalcogenides Based on a Solvable Exciton Model

We develop an analytical procedure to retrieve fundamental material parameters of monolayer transition-metal dichalcogenides from optical and magneto-optical exciton spectra, based on the solvable modified Kratzer model. The proposed retrieval procedure naturally consists of two complementary stages. In the first stage, explicit inversion formulas determine the quasiparticle bandgap, effective screening parameter, and energy scaling factor directly from the experimentally measured energies of the three lowest excitonic states, from which the screening length is subsequently obtained. In the second stage, an analytical expression for the magnetic-field dependence of the exciton energies independently yields the reduced exciton mass, from which the surrounding dielectric constant is then calculated. Once the complete set of material parameters has been retrieved, the framework analytically predicts the diamagnetic coefficients, exciton radii, and complete magnetoexciton spectra without introducing additional fitting parameters or matrix diagonalization. The method is applied to a broad range of experimental samples for WSe$_2$, WS$_2$, MoS$_2$, MoSe$_2$, and MoTe$_2$ monolayers embedded in different dielectric environments. The retrieved material parameters are in good agreement with independent experimental measurements and previous Rytova--Keldysh (RK) calculations, while the predicted excitonic properties accurately reproduce available magneto-optical observations. The proposed analytical theory provides an efficient, physically transparent alternative to conventional numerical fitting procedures and offers an effective tool for the rapid characterization of two-dimensional semiconductors via excitonic spectroscopy.

cond-mat.mtrl-sci↗

arXiv2026

Anisotropic two-dimensional magnetoexciton with exact center-of-mass separation

Excitons in anisotropic two-dimensional (2D) materials, defined by direction-dependent effective masses, are of pronounced interest for their roles in excitonic and magneto-optical phenomena. A perpendicular magnetic field complicates the separation of center-of-mass (c.m.) and relative motions, especially when electron and hole masses are comparable. Conventional theories often employ an approximate c.m. separation using factorized wave functions, modifying magnetic Hamiltonian terms and possibly introducing inaccuracies in magnetoexciton energy predictions. This work develops an exact analytical framework for c.m. and relative motion separation in anisotropic 2D magnetoexcitons, without resorting to the stationary-c.m. approximation. Starting from the full electron-hole Hamiltonian in a homogeneous magnetic field, the formalism uses the conserved pseudomomentum to derive a relative-motion Hamiltonian, revealing new anisotropy-dependent couplings and magnetic coefficients absent in approximate models. The resulting Schrödinger equation is treated via the Feranchuk-Komarov operator method and Levi-Civita transformation, allowing non-perturbative, systematically convergent solutions. Application to monolayer black phosphorus and titanium trisulfide, both freestanding and encapsulated in hexagonal boron nitride, yields magnetoexciton energies, diamagnetic coefficients, and probability densities for the ten lowest states across considerable magnetic-field ranges. The results demonstrate the significant influence of anisotropy-dependent coupling on magnetic response in systems with strong mass anisotropy. This formalism is generalizable to other anisotropic 2D semiconductors, establishing a foundation for advanced magneto-optical studies.

cond-mat.mes-hall↗

arXiv2026

Neuro Symbolic Knowledge Reasoning for Procedural Video Question Answering

In this work we present Knowledge Module Learning (KML) to understand and reason over procedural tasks that requires models to learn structured and compositional procedural knowledge. KML is a neurosymbolic framework that learns relation categories within a knowledge graph as neural knowledge modules and composes them into executable reasoning programs generated by large language models (LLMs). Each module encodes a specific procedural relation capturing how each entity type such as tools are related to steps, purpose of each tool, and steps of each task. Given a question conditioned on a task shown in a video, then KML performs multistep reasoning with transparent, traceable intermediate states. Our theoretical analysis demonstrated two desired properties of KML. KML satisfy strong optimal conditions for modelling KG relations as neural mappings, providing strong foundations for generalizable procedural reasoning. It also shows a bound on the expected error when it performs multistep reasoning. To evaluate this model, we construct a large procedural knowledge graph (PKG) consisting of diverse instructional domains by integrating the COIN instructional video dataset, and COIN ontology, commonsense relations from ConceptNet, and structured extractions from LLMs, followed by expert verification. We then generate question and answer pairs by applying graph traversal templates over the PKG, constructing the PKR-QA benchmark for procedural knowledge reasoning. Experiments show that KML improves structured reasoning performance while providing interpretable step-by-step traces, outperforming LLM-only and black-box neural baselines. Code is publicly available at https://github.com/LUNAProject22/KML.

cs.CV↗

arXiv2025

Harnessing Large Language Models for Scientific Novelty Detection

In an era of exponential scientific growth, identifying novel research ideas is crucial and challenging in academia. Despite potential, the lack of an appropriate benchmark dataset hinders the research of novelty detection. More importantly, simply adopting existing NLP technologies, e.g., retrieving and then cross-checking, is not a one-size-fits-all solution due to the gap between textual similarity and idea conception. In this paper, we propose to harness large language models (LLMs) for scientific novelty detection (ND), associated with two new datasets in marketing and NLP domains. To construct the considerate datasets for ND, we propose to extract closure sets of papers based on their relationship, and then summarize their main ideas based on LLMs. To capture idea conception, we propose to train a lightweight retriever by distilling the idea-level knowledge from LLMs to align ideas with similar conception, enabling efficient and accurate idea retrieval for LLM novelty detection. Experiments show our method consistently outperforms others on the proposed benchmark datasets for idea retrieval and ND tasks. Codes and data are available at https://anonymous.4open.science/r/NoveltyDetection-10FB/.

cs.CL↗

arXiv2025

Keyword-driven Retrieval-Augmented Large Language Models for Cold-start User Recommendations

Recent advancements in Large Language Models (LLMs) have shown significant potential in enhancing recommender systems. However, addressing the cold-start recommendation problem, where users lack historical data, remains a considerable challenge. In this paper, we introduce KALM4Rec (Keyword-driven Retrieval-Augmented Large Language Models for Cold-start User Recommendations), a novel framework specifically designed to tackle this problem by requiring only a few input keywords from users in a practical scenario of cold-start user restaurant recommendations. KALM4Rec operates in two main stages: candidates retrieval and LLM-based candidates re-ranking. In the first stage, keyword-driven retrieval models are used to identify potential candidates, addressing LLMs' limitations in processing extensive tokens and reducing the risk of generating misleading information. In the second stage, we employ LLMs with various prompting strategies, including zero-shot and few-shot techniques, to re-rank these candidates by integrating multiple examples directly into the LLM prompts. Our evaluation, using a Yelp restaurant dataset with user reviews from three English-speaking cities, shows that our proposed framework significantly improves recommendation quality. Specifically, the integration of in-context instructions with LLMs for re-ranking markedly enhances the performance of the cold-start user recommender system.

cs.IR↗

arXiv2020

Structured Self-Attention Weights Encode Semantics in Sentiment Analysis

Neural attention, especially the self-attention made popular by the Transformer, has become the workhorse of state-of-the-art natural language processing (NLP) models. Very recent work suggests that the self-attention in the Transformer encodes syntactic information; Here, we show that self-attention scores encode semantics by considering sentiment analysis tasks. In contrast to gradient-based feature attribution methods, we propose a simple and effective Layer-wise Attention Tracing (LAT) method to analyze structured attention weights. We apply our method to Transformer models trained on two tasks that have surface dissimilarities, but share common semantics---sentiment analysis of movie reviews and time-series valence prediction in life story narratives. Across both tasks, words with high aggregated attention weights were rich in emotional semantics, as quantitatively validated by an emotion lexicon labeled by human annotators. Our results show that structured attention weights encode rich semantics in sentiment analysis, and match human interpretations of semantics.

cs.CL↗

arXiv2020

Injecting Prior Knowledge into Image Caption Generation

Automatically generating natural language descriptions from an image is a challenging problem in artificial intelligence that requires a good understanding of the visual and textual signals and the correlations between them. The state-of-the-art methods in image captioning struggles to approach human level performance, especially when data is limited. In this paper, we propose to improve the performance of the state-of-the-art image captioning models by incorporating two sources of prior knowledge: (i) a conditional latent topic attention, that uses a set of latent variables (topics) as an anchor to generate highly probable words and, (ii) a regularization technique that exploits the inductive biases in syntactic and semantic structure of captions and improves the generalization of image captioning models. Our experiments validate that our method produces more human interpretable captions and also leads to significant improvements on the MSCOCO dataset in both the full and low data regimes.

cs.CL↗

arXiv2019

A Multimodal LSTM for Predicting Listener Empathic Responses Over Time

People naturally understand the emotions of-and often also empathize with-those around them. In this paper, we predict the emotional valence of an empathic listener over time as they listen to a speaker narrating a life story. We use the dataset provided by the OMG-Empathy Prediction Challenge, a workshop held in conjunction with IEEE FG 2019. We present a multimodal LSTM model with feature-level fusion and local attention that predicts empathic responses from audio, text, and visual features. Our best-performing model, which used only the audio and text features, achieved a concordance correlation coefficient (CCC) of 0.29 and 0.32 on the Validation set for the Generalized and Personalized track respectively, and achieved a CCC of 0.14 and 0.14 on the held-out Test set. We discuss the difficulties faced and the lessons learnt tackling this challenge.

cs.CL↗

arXiv2017

Phase diagram of thick ribbons in a bad solvent

Ribbons are topological objects of biological and technological importance. Here, we study the folding of thick ribbons with hydrophobic surfaces in a bad solvent in regimes in which either the ribbon's thickness or the solvent molecule size is not vanishingly small compared to the ribbon's width. Extensive Monte Carlo simulations show that ribbons of various lengths and with a small stiffness adopt several distinct configurations as the ground state that include rolled (Archimedean spiral), curled, twisted and globule conformations. Analytic and numerical calculations based on the consideration of putative ground states lead to phase diagrams that qualitatively agree with the simulation results. A symmetry breaking of the planar rolled configuration in favor of the elongated twisted and the globular ribbons is observed on increasing the solvent size. Interestingly, the twisted ribbon is found as the ground state in the absence of any energetic preference for twisting. We show that the twist of the DNA double helix structure can be stabilized when modeled as a hydrophobic thick ribbon even in the limit of vanishing solvent size.

cond-mat.soft↗

arXiv2016

Theory of Liquid Crystal Elastomers: From Polymer Physics to Differential Geometry

In liquid crystal elastomers, the orientational order of liquid crystals is coupled with elastic distortions of crosslinked polymer networks. Previous theoretical research has described these materials through two different approaches: a neoclassical theory based on the liquid crystal director and the deformation tensor, and a geometric elasticity theory based on the difference between the actual metric tensor and a reference metric. Here, we connect those two approaches using a formalism based on differential geometry. Through this connection, we determine how both the director and the geometry respond to a change of temperature.

cond-mat.soft↗