arXiv subjects

Minhyuk Kim

Publications and source records attributed to Minhyuk Kim.

At least 19 recordsLinked to original sources

arXiv2026

Vitrification-Devitrification Enables Tunable Photonic and Gas Sorption Properties of Zeolitic Imidazolate Frameworks

Zeolitic imidazolate framework (ZIF) glasses represent an emerging family of melt-quenched glasses, which exhibit immense potential for applications in gas separation, energy storage, and optics. However, their intrinsic porosity remains elusive due to the inherent challenges in resolving their disordered atomistic structures. Here, we systematically investigate the porosity of ZIF-4 and ZIF-62 crystals and their corresponding glasses. CO2 sorption at 195 K enables quantitative assessment of microporosity in both crystalline and glassy states, allowing the accessible micropore volume of the ZIF glasses to be determined. Moreover, establishing a direct relationship between photonic properties and structural porosity in Zn-based ZIF glasses remains challenging. Here we demonstrate striking broadband blue-light emission from ZIF-4 glass annealed under optimized conditions. A pronounced red shift is observed when increasing the annealing temperature above the glass transition temperature. By correlating the evolution of photoluminescence with structural porosity, we reveal the interplay between the photonic and gas sorption properties of ZIF glasses. These findings provide new insights into the structure-property relationships of ZIF glasses and offer a pathway toward the rational design of multifunctional MOF glasses.

cond-mat.mtrl-sci↗

arXiv2026

Language Chain in Alignment: Cross-lingual Ranking Preference Optimization

The alignment of Large Language Models heavily relies on English-centric high-quality preference data, which often leads to suboptimal performance in other languages. In this paper, we propose Cross-lingual Ranking Preference Optimization~(CRPO), a novel framework that leverages robust preference knowledge from English to facilitate preference alignment in the target language. We design a hierarchical structure within parallel preference pairs across the target language and English to jointly optimize intra- and inter-lingual preferences, thereby enhancing language adaptation and output quality. Building on the LambdaLoss framework, CRPO goes beyond the binary comparison based optimization by providing a relative ranking signal across multiple candidate responses. Our experiments across five languages with varying resource scales demonstrate that CRPO consistently outperforms standard approaches in both instruction-following and knowledge utilization capability. Notably, the robust performance gains observed across various weighting schemes further validate the empirical effectiveness of our hierarchical design in a multilingual setup. Furthermore, our findings highlight that CRPO significantly improves both reward margins and the log-probability of desirable responses, contributing to a more stable preference manifold for cross-lingual alignment. Our code is available at https://github.com/dltmddbs100/CRPO.

cs.CL↗

arXiv2026

Beyond Penalizing Mistakes: Stabilizing Efficiency Training in Large Reasoning Models via Adaptive Correct-Only Rewards

Training large language models to reason efficiently is a critical challenge. While integrating length-penalizing rewards into Group Relative Policy Optimization (GRPO) aims to reduce verbosity, it frequently triggers reward collapse, severely degrading reasoning capabilities. Through a systematic evaluation of various reward configurations, we identify the root mechanism: GRPO's group normalization creates divergent advantages when incorrect answers receive continuous length penalties. Consequently, methods penalizing the length of incorrect answers are structurally prone to collapse under sustained optimization. Furthermore, restricting penalties exclusively to correct answers avoids this primary failure, but leaves the model susceptible to a stochastic collapse driven by response over-compression. To robustly prevent both failure modes, we propose ACOER (Adaptive Correct-Only Efficiency Reward). ACOER eliminates the structural penalty loop by isolating brevity bonuses to correct completions and prevents stochastic compression via dynamic budget normalization and control-loop penalty adjustments. Evaluated across diverse mathematical reasoning benchmarks, ACOER improves overall accuracy compared to the base model while reducing token generation by over 60%, establishing a fundamentally stable approach for efficiency-aware optimization.

cs.AI↗

arXiv2026

CLEAR: Cross-Lingual Enhancement in Alignment via Reverse-training

Existing multilingual embedding models often encounter challenges in cross-lingual scenarios due to imbalanced linguistic resources and less consideration of cross-lingual alignment during training. Although standardized contrastive learning approaches for cross-lingual adaptation are widely adopted, they may struggle to capture fundamental alignment between languages and degrade performance in well-aligned languages such as English. To address these challenges, we propose Cross-Lingual Enhancement in Retrieval via Reverse-training (CLEAR), a novel loss function utilizing a reverse training scheme to improve retrieval performance across diverse cross-lingual retrieval scenarios. CLEAR leverages an English passage as a bridge to strengthen alignments between the target language and English, ensuring robust performance in the cross-lingual retrieval task. Our extensive experiments demonstrate that CLEAR achieves notable improvements in cross-lingual scenarios, with gains up to 15%, particularly in low-resource languages, while minimizing performance degradation in English. Furthermore, our findings highlight that CLEAR offers promising effectiveness even in multilingual training, suggesting its potential for broad application and scalability. We release the code at https://github.com/dltmddbs100/CLEAR.

cs.CL↗

arXiv2026

Enhanced Maximum Independent Set Preparation with Rydberg Atoms Guided by the Spectral Gap

Adiabatic quantum computation with Rydberg atoms provides a natural route for solving combinatorial optimization problems such as the maximum independent set (MIS). However, its performance is fundamentally limited by the reduction of the spectral gap with increasing system size and connectivity, which induces population leakage from the ground state during finite-time evolution. Here we introduce the Adjusted Detuning for Ground-Energy Leakage Blockade (ADGLB), a spectral-gap-guided schedule engineering method that modifies the laser detuning profile to suppress leakage without introducing additional Hamiltonian terms or iterative optimization loops. We experimentally benchmark ADGLB on a quasi-one-dimensional chain of $N=10$ atoms, and the MIS preparation probability increases substantially compared with the standard adiabatic schedule. Furthermore, we show that the schedule optimized for smaller instances can be directly applied to larger two-dimensional triangular lattices with $N=25$ and $N=37$. With a small heuristic offset, the method also remains effective for instances with higher hardness parameters. These findings demonstrate that spectral-gap-guided schedule engineering offers a scalable and hardware-efficient strategy for enhancing adiabatic quantum optimization on neutral-atom platforms.

quant-ph↗

arXiv2026

What Users Leave Unsaid: Under-Specified Queries Limit Vision-Language Models

Current vision-language benchmarks predominantly feature well-structured questions with clear, explicit prompts. However, real user queries are often informal and underspecified. Users naturally leave much unsaid, relying on images to convey context. We introduce HAERAE-Vision, a benchmark of 653 real-world visual questions from Korean online communities (0.76% survival from 86K candidates), each paired with an explicit rewrite, yielding 1,306 query variants in total. Evaluating 39 VLMs, we find that even state-of-the-art models (GPT-5, Gemini 2.5 Pro) achieve under 50% on the original queries. Crucially, query explicitation alone yields 8 to 22 point improvements, with smaller models benefiting most. We further show that even with web search, under-specified queries underperform explicit queries without search, revealing that current retrieval cannot compensate for what users leave unsaid. Our findings demonstrate that a substantial portion of VLM difficulty stem from natural query under-specification instead of model capability, highlighting a critical gap between benchmark evaluation and real-world deployment.

cs.CV↗

arXiv2025

Pushing on Multilingual Reasoning Models with Language-Mixed Chain-of-Thought

Recent frontier models employ long chain-of-thought reasoning to explore solution spaces in context and achieve stonger performance. While many works study distillation to build smaller yet capable models, most focus on English and little is known about language-specific reasoning. To bridge this gap, we first introduct **Language-Mixed CoT**, a reasoning schema that switches between English and a target language, using English as an anchor to excel in reasoning while minimizing translation artificats. As a Korean case study, we curate **Yi-Sang**: 5.79M native-Korean prompts from web Q&A, exams, STEM, and code; 3.7M long reasoning traces generated from Qwen3-32B; and a targeted 260k high-yield subset. We train ninve models (4B-35B) across six families (Qwen2.5, Llama-3.1, Gemma-3, etc). Our best model, **KO-REAson-35B**, achieves state-of-the-art performance, with the highest overall average score (64.0 \pm 25), ranking first on 5/9 benchmarks and second on the remainder. Samller and mid-sized models also benefit substantially, with an average improvement of +18.6 points across teh evaluated nine benchmarks. Ablations show **Language-Mixed CoT** is more effective than monolingual CoT, also resulting in cross-lingual and mult-modal performance gains. We release our data-curation pipeline, evaluation system, datasets, and models to advance research on language-specific reasoning. Data and model collection: https://huggingface.co/KOREAson.

cs.CL↗

arXiv2025

MAQCY: Modular Atom-Array Quantum Computing with Space-Time Hybrid Multiplexing

We present a modular atom-array quantum computing architecture with space-time hybrid multiplexing (MAQCY), a dynamic optical tweezer-based protocol for fully connected and scalable universal quantum computation. By extending the concept of globally controlled static dual-species Rydberg atom wires [1], we develop an entirely new approach using Q-Pairs, which consist of globally controlled and temporally multiplexed dual-species Rydberg blockaded atom and superatom pairs. Space-time hybrid multiplexing of Q-Pairs achieves O(N) linear scaling in the number of required physical qubits, while preserving coherence and mitigating circuit-depth limitations through in-situ atom replacement. To demonstrate MAQCY's versatility, we implement a three-qubit quantum Fourier transform using only global operations and atom transport. We also propose a concrete implementation using ytterbium isotopes, paving the way toward large-scale, fault-tolerant quantum computing.

quant-ph↗

arXiv2025

Benchmark Profiling: Mechanistic Diagnosis of LLM Benchmarks

Large Language Models are commonly judged by their scores on standard benchmarks, yet such scores often overstate real capability since they mask the mix of skills a task actually demands. For example, ARC is assumed to test reasoning, while HellaSwag is designed to evaluate commonsense. However, we lack a systematic way to verify if these benchmarks actually measure these labels. We introduce Benchmark Profiling, a diagnostic framework that decomposes benchmark performance into ten cognitively grounded abilities. The method combines gradient-based importance scoring with targeted parameter ablation to compute an Ability Impact Score (AIS) that quantifies how much each ability contributes to a model's success on a given benchmark. Profiling three instruction-tuned models across ten widely used benchmarks yields four key findings: (i) most benchmarks draw on several abilities rather than one, (ii) datasets with similar labels rely on distinct ability mixtures, (iii) code-generation benchmarks reward broad, multi-skill improvement and thus show only modest gains from narrow domain-specific fine-tuning, and (iv) abilities irrelevant to the task could negatively affect performance. Benchmark Profiling therefore explains why performance gains do not always translate into user-perceived competence and offers a transparent tool for benchmark audit and model interpretability.

cs.CL↗

arXiv2025

TORSO: Template-Oriented Reasoning Towards General Tasks

The approaches that guide Large Language Models (LLMs) to emulate human reasoning during response generation have emerged as an effective method for enabling them to solve complex problems in a step-by-step manner, thereby achieving superior performance. However, most existing approaches using few-shot prompts to generate responses heavily depend on the provided examples, limiting the utilization of the model's inherent reasoning capabilities. Moreover, constructing task-specific few-shot prompts is often costly and may lead to inconsistencies across different tasks. In this work, we introduce Template-Oriented Reasoning (TORSO), which elicits the model to utilize internal reasoning abilities to generate proper responses across various tasks without the need for manually crafted few-shot examples. Our experimental results demonstrate that TORSO achieves strong performance on diverse LLMs benchmarks with reasonable rationales.

cs.AI↗

arXiv2025

Enhancing Automatic Term Extraction with Large Language Models via Syntactic Retrieval

Automatic Term Extraction (ATE) identifies domain-specific expressions that are crucial for downstream tasks such as machine translation and information retrieval. Although large language models (LLMs) have significantly advanced various NLP tasks, their potential for ATE has scarcely been examined. We propose a retrieval-based prompting strategy that, in the few-shot setting, selects demonstrations according to \emph{syntactic} rather than semantic similarity. This syntactic retrieval method is domain-agnostic and provides more reliable guidance for capturing term boundaries. We evaluate the approach in both in-domain and cross-domain settings, analyzing how lexical overlap between the query sentence and its retrieved examples affects performance. Experiments on three specialized ATE benchmarks show that syntactic retrieval improves F1-score. These findings highlight the importance of syntactic cues when adapting LLMs to terminology-extraction tasks.

cs.CL↗

arXiv2025

Prospects to bypass nonlocal phenomena in metals using phonon-polaritons

Electromagnetic design relies on an accurate understanding of light-matter interactions, yet often overlooks electronic length scales. Under extreme confinement, this omission can lead to nonclassical effects, such as nonlocal response. Here, we use mid-infrared phonon-polaritons in hexagonal boron nitride (hBN) screened by monocrystalline gold flakes to push the limits of nanolight confinement unobstructed by nonlocal phenomena, even when the polariton phase velocity approaches the Fermi velocities of electrons in gold. We employ near-field imaging to probe polaritons in nanometre-thin crystals of hBN on gold and extract their complex propagation constant, observing effective indices exceeding 90. We further show the importance of sample characterisation by revealing a thin low-index interfacial layer naturally forming on monocrystalline gold. Our experiments address a fundamental limitation posed by nonlocal effects in van der Waals heterostructures and outline a pathway to bypass their impact in high-confinement regimes.

physics.optics↗

arXiv2024

Exploring Coding Spot: Understanding Parametric Contributions to LLM Coding Performance

Large Language Models (LLMs) have demonstrated notable proficiency in both code generation and comprehension across multiple programming languages. However, the mechanisms underlying this proficiency remain underexplored, particularly with respect to whether distinct programming languages are processed independently or within a shared parametric region. Drawing an analogy to the specialized regions of the brain responsible for distinct cognitive functions, we introduce the concept of Coding Spot, a specialized parametric region within LLMs that facilitates coding capabilities. Our findings identify this Coding Spot and show that targeted modifications to this subset significantly affect performance on coding tasks, while largely preserving non-coding functionalities. This compartmentalization mirrors the functional specialization observed in cognitive neuroscience, where specific brain regions are dedicated to distinct tasks, suggesting that LLMs may similarly employ specialized parameter regions for different knowledge domains.

cs.CL↗

arXiv2023

A Rydberg-atom approach to the integer factorization problem

The task of factoring integers poses a significant challenge in modern cryptography, and quantum computing holds the potential to efficiently address this problem compared to classical algorithms. Thus, it is crucial to develop quantum computing algorithms to address this problem. This study introduces a quantum approach that utilizes Rydberg atoms to tackle the factorization problem. Experimental demonstrations are conducted for the factorization of small composite numbers such as $6 = 2 \times 3$, $15 = 3 \times 5$, and $35 = 5 \times 7$. This approach involves employing Rydberg-atom graphs to algorithmically program binary multiplication tables, yielding many-body ground states that represent superpositions of factoring solutions. Subsequently, these states are probed using quantum adiabatic computing. Limitations of this method are discussed, specifically addressing the scalability of current Rydberg quantum computing for the intricate computational problem.

quant-ph↗

arXiv2023

Quantum Computing Dataset of Maximum Independent Set Problem on King's Lattice of over Hundred Rydberg Atoms

Finding the maximum independent set (MIS) of a large-size graph is a nondeterministic polynomial-time (NP)-complete problem not efficiently solvable with classical computations. Here, we present a set of quantum adiabatic computing data of Rydberg-atom experiments performed to solve the MIS problem of up to 141 atoms randomly arranged on the King's lattice. A total of 582,916 events of Rydberg-atom measurements are collected for experimental MIS solutions of 733,853 different graphs. We provide the raw image data along with the entire binary determinations of the measured many-body ground states and the classified graph data, to offer bench-mark testing and advanced data-driven analyses for validation of the performance and system improvements of the Rydberg-atom approach.

quant-ph↗

arXiv2023

Rydberg-atom graphs for quadratic unconstrained binary optimization problems

There is a growing interest in harnessing the potential of the Rydberg-atom system to address complex combinatorial optimization challenges. Here we present an experimental demonstration of how the quadratic unconstrained binary optimization (QUBO) problem can be effectively addressed using Rydberg-atom graphs. The Rydberg-atom graphs are configurations of neutral atoms organized into mathematical graphs, facilitated by programmable optical tweezers, and designed to exhibit many-body ground states that correspond to the maximum independent set (MIS) of their respective graphs. We have developed four elementary Rydberg-atom subgraph components, not only to eliminate the need of local control but also to be robust against interatomic distance errors, while serving as the building blocks sufficient for formulating generic QUBO graphs. To validate the feasibility of our approach, we have conducted a series of Rydberg-atom experiments selected to demonstrate proof-of-concept operations of these building blocks. These experiments illustrate how these components can be used to programmatically encode the QUBO problems to Rydberg-atom graphs and, by measuring their many-body ground states, how their QUBO solutions are determined subsequently.

quant-ph↗

arXiv2023

Exploring the impact of graph locality for the resolution of MIS with neutral atom devices

In the past years, many quantum algorithms have been proposed to tackle hard combinatorial problems. In particular, the Maximum Independent Set (MIS) is a known NP-hard problem that can be naturally encoded in Rydberg atom arrays. By representing a graph with an ensemble of neutral atoms one can leverage Rydberg dynamics to naturally encode the constraints and the solution to MIS. However, the classes of graphs that can be directly mapped ``vertex-to-atom" on standard devices with 2D capabilities are currently limited to Unit-Disk graphs. In this setting, the inherent spatial locality of the graphs can be leveraged by classical polynomial-time approximation schemes (PTAS) that guarantee an $ε$-approximate solution. In this work, we build upon recent progress made for using 3D arrangements of atoms to embed more complex classes of graphs. We report experimental and theoretical results which represent important steps towards tackling combinatorial tasks on quantum computers for which no classical efficient $\varepsilon$-approximation scheme exists.

quant-ph↗

arXiv2023

Quantum Programming of the Satisfiability Problem with Rydberg Atom Graphs

Finding a quantum computing method to solve nondeterministic polynomial time (NP)-complete problems is currently of paramount importance in quantum information science. Here an experiment is presented to demonstrate the use of Rydberg atoms to solve (i.e., to program and obtain the solution of) the satisfiability (3-SAT) problem, which is the prototypical NP-complete problem allowing general programming of all NP problems. Boolean expressions of the 3-SAT problem are programmed with the blockade interactions of Rydberg atom graphs and their many-body ground states are experimentally obtained, to determine the satisfiabilities of the given 3-SAT problem instances quantum mechanically.

quant-ph↗