SearcharxivSearch

arXiv subjects

Daeyoung Park

Publications and source records attributed to Daeyoung Park.

4 recordsLinked to original sources

FinRED: An Expert-Guided Benchmark Generation and Evaluation Framework for Financial LLM Red-Teaming

Existing safety benchmarks target general adversarial scenarios but miss finance-specific risks. Financial LLMs face regulatory compliance violations, fraud facilitation, and systemic trust erosion that require targeted evaluation. We introduce FinRED, an expert-guided red-teaming framework for financial LLM safety evaluation developed with financial experts. FinRED uses a novel two-level taxonomy mapping global standards (e.g., FATF and EU DORA) to threats ranging from regulatory evasion to complex fraud, integrated with a scalable pipeline that converts real financial documents into context-rich red-teaming Behavioral Prompts (seeds) through an expert-defined schema. Rigorous expert validation confirms seed plausibility and realism for meaningful LLM safety evaluation. We also provide an expert-validated, finance-specific rubric that goes beyond disclaimer checks, aligns more closely with human experts than static one-size-fits-all rubrics, and reduces critical false negatives from 28 to 12. Aligned with internationally adopted risk-management and information-security standards (e.g., ISO/IEC 27001), FinRED is deployed in South Korea's Financial Security Institute (FSI) regulatory sandbox for generative AI security evaluation in real financial services. To mitigate dual-use risks, the dataset, generation pipeline, prompt template, and evaluation framework are gated for qualified researchers at https://github.com/selectstar-ai/FinRED-paper and https://huggingface.co/datasets/datumo/FinRED.

cs.CR

VOMTC: Vision Objects for Millimeter and Terahertz Communications

Recent advances in sensing and computer vision (CV) technologies have opened the door for the application of deep learning (DL)-based CV technologies in the realm of 6G wireless communications. For the successful application of this emerging technology, it is crucial to have a qualified vision dataset tailored for wireless applications (e.g., RGB images containing wireless devices such as laptops and cell phones). An aim of this paper is to propose a large-scale vision dataset referred to as Vision Objects for Millimeter and Terahertz Communications (VOMTC). The VOMTC dataset consists of 20,232 pairs of RGB and depth images obtained from a camera attached to the base station (BS), with each pair labeled with three representative object categories (person, cell phone, and laptop) and bounding boxes of the objects. Through experimental studies of the VOMTC datasets, we show that the beamforming technique exploiting the VOMTC-trained object detector outperforms conventional beamforming techniques.

cs.NI

Combination of Optical Transitions of Polarons with Rashba Effect in Methylammonium Lead Tri-halide Perovskites under High Magnetic Fields

We investigate photoluminescence (PL) transitions of MAPbX$_{3}$ (X = I, Br and Cl) organic-inorganic hybrid perovskite single crystals under magnetic fields of up to 60 T. In these materials, sharp free-exciton transition peaks emerge at a low temperature (4.2 K). Under strong magnetic fields, the free-exciton PL transitions of three different halogens show dramatic differences. The free-exciton transitions of the MAPbCl$_{3}$ crystal undergo negative energy shifts, while those of the MAPbBr$_{3}$ crystal show normal diamagnetic shifts. To obtain the variation from Cl to Br, we attempt to measure PL transitions of MAPbCl$_{x}$Br$_{3-x}$. For MAPbI$_{3}$, the transition-energy shifts for both $σ^{+}$ and $σ^{-}$ transitions at 4.2 K exhibit a power-law dependence on the magnetic field. Such inconsistent magnetic-field effects on different halogens make it difficult to understand the transition-energy behavior through a unified model. We propose a possible mechanism for the field effects that is based on a combination of the Rashba effect induced by strong spin-orbit coupling and the polaron effect caused by the polar nature of the inorganic elements.

cond-mat.mtrl-sci

Hard Fairness Versus Proportional Fairness in Wireless Communications: The Multiple-Cell Case

We consider the uplink of a cellular communication system with $K$ users per cell and infinite base stations equally spaced on a line. The system is conventional, i.e., it does not make use of joint cell-site processing. A hard fairness (HF) system serves all users with the same rate in any channel state. In contrast, a system based on proportional fairness serves the users with variable instantaneous rates depending on their channel state. We compare these two options in terms of the system spectral efficiency \textsf{C} (bit/s/Hz) versus $E_b/N_0$. Proportional fair scheduling (PFS) performs generally better than the more restrictive HF system in the regime of low to moderate SNR, but for high SNR an optimized HF system achieves throughput comparable to that of PFS system for finite $K$. The hard-fairness system is interference limited. We characterize this limit and validate a commonly used simplified model that treats outer cell interference power as proportional to the in-cell total power and we analytically characterize the proportionality constant. In contrast, the spectral efficiency of PFS can grow unbounded for $K \to \infty$ thanks to the multiuser diversity effect. We also show that partial frequency/time reuse can mitigate the throughput penalty of the HF system, especially at high SNR.

cs.IT