arXiv subjects

Jiwoo Choi

Publications and source records attributed to Jiwoo Choi.

8 recordsLinked to original sources

arXiv2026

Anchoring LLM Gender Bias to Human Baselines: A Cross-Lingual Audit

We audit six large language models (LLMs) for gender stereotyping across English, Korean, Chinese, and Japanese. Three were developed primarily for English-language use (Claude, GPT, Gemini) and three for East Asian use (DeepSeek, Syn-Pro, HyperCLOVA X). We adopt the HEXACO-100 personality inventory and anchor each model against a cross-cultural human dataset spanning 48 countries to ask not whether LLMs are biased, but how far their gender attributions drift from the populations they are deployed among. Our findings show that their stereotyping spans a range roughly 2.5 times wider than the entire cross-country range found in humans, and the effect can compound across languages. One English-centric model, prompted in Korean, reached 5 times the local baseline, even when the prompt stated the candidate had already been hired, which often dampens human stereotyping. To characterize such behaviors without ranking them, we introduce a four-pattern framework -- concordance, suppression, reorganization, and amplification -- across 24 (model x language) cells. Item-level analysis reveals that translation does not just rescale stereotypes, but changes the attributes tied to it, hiding significant rearrangement under the surface while appearing well-calibrated. Our results ultimately suggest that no single debiasing pipeline is likely to address bias evenly across linguistic boundaries.

cs.CL↗

arXiv2026

Narrative Landscape: Mapping Narrative Dispositions Across LLMs

This study proposes a quantitative framework for profiling LLM dispositions as stable, model-specific regularities in output under repeated, controlled elicitation. Using a structured narrative constraint-selection task administered across six frontier models and three instruction types, we operationalize disposition through two dimensions: "consistency", measured as cross-replication selection overlap via Jaccard similarity, and "diversity", measured as dispersion across options via the inverse Simpson index. We further introduce Narrative Landscape, a PCA-based visualization that maps each model's selection profile into a shared space for direct comparison. Results reveal a clear rigidity-exploration spectrum across model families and show that instruction types shift the geometry of selection spaces even when scalar metrics appear similar, indicating that comparable scores can mask qualitatively distinct selection topologies.

cs.CL↗

arXiv2025

Style over Story: Measuring LLM Narrative Preferences via Structured Selection

We introduce a constraint-selection-based experiment design for measuring narrative preferences of Large Language Models (LLMs). This design offers an interpretable lens on LLMs' narrative selection behavior. We developed a library of 200 narratology-grounded constraints and prompted selections from six LLMs under three different instruction types: basic, quality-focused, and creativity-focused. Findings demonstrate that models consistently prioritize Style over narrative content elements like Event, Character, and Setting. Style preferences remain stable across models and instruction types, whereas content elements show cross-model divergence and instructional sensitivity. These results suggest that LLMs have latent narrative preferences, which should inform how the NLP community evaluates and deploys models in creative domains.

cs.CL↗

arXiv2025

ORIDa: Object-centric Real-world Image Composition Dataset

Object compositing, the task of placing and harmonizing objects in images of diverse visual scenes, has become an important task in computer vision with the rise of generative models. However, existing datasets lack the diversity and scale required to comprehensively explore real-world scenarios. We introduce ORIDa (Object-centric Real-world Image Composition Dataset), a large-scale, real-captured dataset containing over 30,000 images featuring 200 unique objects, each of which is presented across varied positions and scenes. ORIDa has two types of data: factual-counterfactual sets and factual-only scenes. The factual-counterfactual sets consist of four factual images showing an object in different positions within a scene and a single counterfactual (or background) image of the scene without the object, resulting in five images per scene. The factual-only scenes include a single image containing an object in a specific context, expanding the variety of environments. To our knowledge, ORIDa is the first publicly available dataset with its scale and complexity for real-world image composition. Extensive analysis and experiments highlight the value of ORIDa as a resource for advancing further research in object compositing.

cs.CV↗

arXiv2024

A 2-step Framework for Automated Literary Translation Evaluation: Its Promises and Pitfalls

In this work, we propose and evaluate the feasibility of a two-stage pipeline to evaluate literary machine translation, in a fine-grained manner, from English to Korean. The results show that our framework provides fine-grained, interpretable metrics suited for literary translation and obtains a higher correlation with human judgment than traditional machine translation metrics. Nonetheless, it still fails to match inter-human agreement, especially in metrics like Korean Honorifics. We also observe that LLMs tend to favor translations generated by other LLMs, and we highlight the necessity of developing more sophisticated evaluation methods to ensure accurate and culturally sensitive machine translation of literary works.

cs.CL↗

arXiv2024

MaTableGPT: GPT-based Table Data Extractor from Materials Science Literature

Efficiently extracting data from tables in the scientific literature is pivotal for building large-scale databases. However, the tables reported in materials science papers exist in highly diverse forms; thus, rule-based extractions are an ineffective approach. To overcome this challenge, we present MaTableGPT, which is a GPT-based table data extractor from the materials science literature. MaTableGPT features key strategies of table data representation and table splitting for better GPT comprehension and filtering hallucinated information through follow-up questions. When applied to a vast volume of water splitting catalysis literature, MaTableGPT achieved an extraction accuracy (total F1 score) of up to 96.8%. Through comprehensive evaluations of the GPT usage cost, labeling cost, and extraction accuracy for the learning methods of zero-shot, few-shot and fine-tuning, we present a Pareto-front mapping where the few-shot learning method was found to be the most balanced solution owing to both its high extraction accuracy (total F1 score>95%) and low cost (GPT usage cost of 5.97 US dollars and labeling cost of 10 I/O paired examples). The statistical analyses conducted on the database generated by MaTableGPT revealed valuable insights into the distribution of the overpotential and elemental utilization across the reported catalysts in the water splitting literature.

cs.CL↗

arXiv2024

Optimal Impact Angle Guidance via First-Order Optimization under Nonconvex Constraints

Most of the optimal guidance problems can be formulated as nonconvex optimization problems, which can be solved indirectly by relaxation, convexification, or linearization. Although these methods are guaranteed to converge to the global optimum of the modified problems, the obtained solution may not guarantee global optimality or even the feasibility of the original nonconvex problems. In this paper, we propose a computational optimal guidance approach that directly handles the nonconvex constraints encountered in formulating the guidance problems. The proposed computational guidance approach alternately solves the least squares problems and projects the solution onto nonconvex feasible sets, which rapidly converges to feasible suboptimal solutions or sometimes to the globally optimal solutions. The proposed algorithm is verified via a series of numerical simulations on impact angle guidance problems under state dependent maneuver vector constraints, and it is demonstrated that the proposed algorithm provides superior guidance performance than conventional techniques.

math.OC↗

arXiv2024

Powered Descent Guidance via First-Order Optimization with Expansive Projection

This paper introduces a first-order method for solving optimal powered descent guidance (PDG) problems, that directly handles the nonconvex constraints associated with the maximum and minimum thrust bounds with varying mass and the pointing angle constraints on thrust vectors. This issue has been conventionally circumvented via lossless convexification (LCvx), which lifts a nonconvex feasible set to a higher-dimensional convex set, and via linear approximation of another nonconvex feasible set defined by exponential functions. However, this approach sometimes results in an infeasible solution when the solution obtained from the higher-dimensional space is projected back to the original space, especially when the problem involves a nonoptimal time of flight. Additionally, the Taylor series approximation introduces an approximation error that grows with both flight time and deviation from the reference trajectory. In this paper, we introduce a first-order approach that makes use of orthogonal projections onto nonconvex sets, allowing expansive projection (ExProj). We show that 1) this approach produces a feasible solution with better performance even for the nonoptimal time of flight cases for which conventional techniques fail to generate achievable trajectories and 2) the proposed method compensates for the linearization error that arises from Taylor series approximation, thus generating a superior guidance solution with less fuel consumption. We provide numerical examples featuring quantitative assessments to elucidate the effectiveness of the proposed methodology, particularly in terms of fuel consumption and flight time. Our analysis substantiates the assertion that the proposed approach affords enhanced flexibility in devising viable trajectories for a diverse array of planetary soft landing scenarios.

math.OC↗