arXiv subjects

Eunseok Lee

Publications and source records attributed to Eunseok Lee.

7 recordsLinked to original sources

arXiv2025

Automating Code Generation for Semiconductor Equipment Control from Developer Utterances with LLMs

Semiconductors form the backbone of modern electronics, with their manufacturing and testing relying on highly specialized equipment and domain-specific programming languages. Equipment languages such as the Algorithmic Pattern Generator (ALPG) are critical for precise hardware control but are challenging to program due to their low-level syntax and steep learning curve. While large language models (LLMs) have shown promise in generating high-level code from natural language, their effectiveness on low-level equipment languages remains limited. To address this, we propose Progressive Knowledge Enhancement (PKE), a novel multi-stage prompting framework that progressively extracts and activates the latent knowledge within LLMs, guiding them from simple to complex examples without extensive fine-tuning. Empirical evaluation on an industrial ALPG dataset shows that PKE significantly outperforms standard prompting and surpasses state-of-the-art methods in generating correct ALPG code, achieving 11.1\% and 15.2\% higher exact match scores compared to the second-best technique. Further analysis of individual components confirms that progressive knowledge extraction based on difficulty enhances accuracy. Our study offer a practical approach to boosting LLM capabilities for specialized low-level programming, supporting greater productivity in semiconductor software development.

cs.SE↗

arXiv2025

Testing SSD Firmware with State Data-Aware Fuzzing: Accelerating Coverage in Nondeterministic I/O Environments

Solid-State Drive (SSD) firmware manages complex internal states, including flash memory maintenance. Due to nondeterministic I/O operations, traditional testing methods struggle to rapidly achieve coverage of firmware code areas that require extensive I/O accumulation. To address this challenge, we propose a state data-aware fuzzing approach that leverages SSD firmware's internal state to guide input generation under nondeterministic I/O conditions and accelerate coverage discovery. Our experiments with an open-source SSD firmware emulator show that the proposed method achieves the same firmware test coverage as a state-of-the-art coverage-based fuzzer (AFL++) while requiring approximately 67% fewer commands, without reducing the number of crashes or hangs detected. Moreover, we extend our experiments by incorporating various I/O commands beyond basic write/read operations to reflect real user scenarios, and we confirm that our strategy remains effective even for multiple types of I/O tests. We further validate the effectiveness of state data-aware fuzzing for firmware testing under I/O environments and suggest that this approach can be extended to other storage firmware or threshold-based embedded systems in the future.

cs.SE↗

arXiv2023

Fault-tolerant resource estimate for quantum chemical simulations: Case study on Li-ion battery electrolyte molecules

We estimate the resources required in the fusion-based quantum computing scheme to simulate electrolyte molecules in Li-ion batteries on a fault-tolerant, photonic quantum computer. We focus on the molecules that can provide practical solutions to industrially relevant problems. Certain fault-tolerant operations require the use of single-qubit "magic states" prepared by dedicated "magic state factories" (MSFs). Producing and consuming magic states in parallel is typically a prohibitively expensive task, resulting in the serial application of fault-tolerant gates. However, for the systems considered, the MSF constitutes a negligible fraction of the total footprint of the quantum computer, allowing for the use of multiple MSFs to produce magic states in parallel. We suggest architectural and algorithmic techniques that can accommodate such a capability. We propose a method to consume multiple magic states simultaneously, which can potentially lead to an order of magnitude reduction in the computational runtime without additional expense in the footprint.

quant-ph↗

arXiv2021

Quantum simulation of electronic structure with a transcorrelated Hamiltonian: improved accuracy with a smaller footprint on the quantum computer

Quantum simulations of electronic structure with a transformed Hamiltonian that includes some electron correlation effects are demonstrated. The transcorrelated Hamiltonian used in this work is efficiently constructed classically, at polynomial cost, by an approximate similarity transformation with an explicitly correlated two-body unitary operator. This Hamiltonian is Hermitian, includes no more than two-particle interactions, and is free of electron-electron singularities. We investigate the effect of such a transformed Hamiltonian on the accuracy and computational cost of quantum simulations by focusing on a widely used solver for the Schrodinger equation, namely the variational quantum eigensolver method, based on the unitary coupled cluster with singles and doubles (q-UCCSD) Ansatz. Nevertheless, the formalism presented here translates straightforwardly to other quantum algorithms for chemistry. Our results demonstrate that a transcorrelated Hamiltonian, paired with extremely compact bases, produces explicitly correlated energies comparable to those from much larger bases. For the chemical species studied here, explicitly correlated energies based on an underlying 6-31G basis had cc-pVTZ quality. The use of the very compact transcorrelated Hamiltonian reduces the number of CNOT gates required to achieve cc-pVTZ quality by up to two orders of magnitude, and the number of qubits by a factor of three.

quant-ph↗

arXiv2021

A Heuristic Quantum-Classical Algorithm for Modeling Substitutionally Disordered Binary Crystalline Materials

Improving the efficiency and accuracy of energy calculations has been of significant and continued interest in the area of materials informatics, a field that applies machine learning techniques to computational materials data. Here, we present a heuristic quantum-classical algorithm to efficiently model and predict the energies of substitutionally disordered binary crystalline materials. Specifically, a quantum circuit that scales linearly in the number of lattice sites is designed and trained to predict the energies of quantum chemical simulations in an exponentially-scaling feature space. This circuit is trained by classical supervised-learning using data obtained from classically-computed quantum chemical simulations. As a part of the training process, we introduce a sub-routine that is able to detect and rectify anomalies in the input data. The algorithm is demonstrated on the complex layer-structured of Li-cobaltate system, a widely-used Li-ion battery cathode material component. Our results shows that the proposed quantum circuit model presents a suitable choice for modelling the energies obtained from such quantum mechanical systems. Furthermore, analysis of the anomalous data provides important insights into the thermodynamic properties of the systems studied.

quant-ph↗

arXiv2019

Increasing the representation accuracy of quantum simulations of chemistry without extra quantum resources

Proposals for near-term experiments in quantum chemistry on quantum computers leverage the ability to target a subset of degrees of freedom containing the essential quantum behavior, sometimes called the active space. This approximation allows one to treat more difficult problems using fewer qubits and lower gate depths than would otherwise be possible. However, while this approximation captures many important qualitative features, it may leave the results wanting in terms of absolute accuracy (basis error) of the representation. In traditional approaches, increasing this accuracy requires increasing the number of qubits and an appropriate increase in circuit depth as well. Here we introduce a technique requiring no additional qubits or circuit depth that is able to remove much of this approximation in favor of additional measurements. The technique is constructed and analyzed theoretically, and some numerical proof of concept calculations are shown. As an example, we show how to achieve the accuracy of a 20 qubit representation using only 4 qubits and a modest number of additional measurements for a simple hydrogen molecule. We close with an outlook on the impact this technique may have on both near-term and fault-tolerant quantum simulations.

quant-ph↗

arXiv2010

Enhancing Ionic Conductivity of Bulk Single Crystal Yttria-Stabilized Zirconia by Tailoring Dopant Distribution

We present an ab-initio based kinetic Monte Carlo model for ionic conductivity in single crystal yttria-stabilized zirconia. Ionic interactions are taken into account by combining density functional theory calculations and the cluster expansion method and are found to be essential in reproducing the effective activation energy observed in experiments. The model predicts that the effective energy barrier can be reduced by 0.15-0.25 eV by arranging the dopant ions into a super-lattice.

cond-mat.mtrl-sci↗