arXiv subjects

Junyeong Park

Publications and source records attributed to Junyeong Park.

At least 37 records · Page 2Linked to original sources

arXiv2025

Diffusion Models Through a Global Lens: Are They Culturally Inclusive?

Text-to-image diffusion models have recently enabled the creation of visually compelling, detailed images from textual prompts. However, their ability to accurately represent various cultural nuances remains an open question. In our work, we introduce CultDiff benchmark, evaluating state-of-the-art diffusion models whether they can generate culturally specific images spanning ten countries. We show that these models often fail to generate cultural artifacts in architecture, clothing, and food, especially for underrepresented country regions, by conducting a fine-grained analysis of different similarity aspects, revealing significant disparities in cultural relevance, description fidelity, and realism compared to real-world reference images. With the collected human evaluations, we develop a neural-based image-image similarity metric, namely, CultDiff-S, to predict human judgment on real and generated images with cultural artifacts. Our work highlights the need for more inclusive generative AI systems and equitable dataset representation over a wide range of cultures.

cs.CV↗

arXiv2025

MrSteve: Instruction-Following Agents in Minecraft with What-Where-When Memory

Significant advances have been made in developing general-purpose embodied AI in environments like Minecraft through the adoption of LLM-augmented hierarchical approaches. While these approaches, which combine high-level planners with low-level controllers, show promise, low-level controllers frequently become performance bottlenecks due to repeated failures. In this paper, we argue that the primary cause of failure in many low-level controllers is the absence of an episodic memory system. To address this, we introduce MrSteve (Memory Recall Steve), a novel low-level controller equipped with Place Event Memory (PEM), a form of episodic memory that captures what, where, and when information from episodes. This directly addresses the main limitation of the popular low-level controller, Steve-1. Unlike previous models that rely on short-term memory, PEM organizes spatial and event-based data, enabling efficient recall and navigation in long-horizon tasks. Additionally, we propose an Exploration Strategy and a Memory-Augmented Task Solving Framework, allowing agents to alternate between exploration and task-solving based on recalled events. Our approach significantly improves task-solving and exploration efficiency compared to existing methods. We will release our code and demos on the project page: https://sites.google.com/view/mr-steve.

cs.LG↗

arXiv2025

Possibility of using 3.3$μ$m PAH luminosity as a molecular gas mass estimator

We present CO(1-0) observations of 50 star-forming galaxies at 0.01<z<0.35, for which 3.3$\,μ$m PAH emission flux or its upper limit is available. A scaling relation between 3.3$\,μ$m PAH luminosity and CO(1-0) luminosity is established covering ~2 orders of magnitude in total IR luminosity and CO luminosity, with a scatter of ~0.23 dex: $\mathrm{log}\,L_\mathrm{3.3}/\mathrm{L}_\odot=(1.00\pm0.07)\times\mathrm{log}\,L_\mathrm{CO(1-0)}^\prime/(\mathrm{K\,km\,s^{-1}\,pc^2})+(-1.10\pm0.70)$. The slope is near unity, allowing the use of a single value of $\langle\mathrm{log}\,(L_\mathrm{3.3}/L_\mathrm{CO(1-0)}^\prime)\rangle=-1.09\pm0.36~[\mathrm{L}_\odot/(\mathrm{K\,km\,s^{-1}\,pc^2})]$ in the conversion between 3.3$\,μ$m PAH and CO luminosities. The variation in the $L_\mathrm{3.3}/L_\mathrm{CO}^\prime$ ratio is not dependent on the galaxy properties, including total IR luminosity, stellar mass, and SFR excess. The total gas mass, estimated using dust-to-gas ratio and dust mass, is correlated with 3.3$\,μ$m PAH luminosity, in line with the prescription using $α_\mathrm{CO}=0.8-4.5$ covering both normal star-forming galaxies and starburst galaxies. AGN-dominated galaxies tend to have a lower $L_\mathrm{3.3}/L_\mathrm{CO}^\prime$ than non-AGN galaxies, which needs to be investigated further with an increased sample size. The established $L_\mathrm{3.3}$-$L_\mathrm{CO}^\prime$ correlation is expected to be applicable to wide-field near-infrared spectrophotometric surveys that allow the detection of 3.3$\,μ$m emission from numerous low-redshift galaxies.

astro-ph.GA↗

arXiv2025

When Tom Eats Kimchi: Evaluating Cultural Bias of Multimodal Large Language Models in Cultural Mixture Contexts

In a highly globalized world, it is important for multi-modal large language models (MLLMs) to recognize and respond correctly to mixed-cultural inputs. For example, a model should correctly identify kimchi (Korean food) in an image both when an Asian woman is eating it, as well as an African man is eating it. However, current MLLMs show an over-reliance on the visual features of the person, leading to misclassification of the entities. To examine the robustness of MLLMs to different ethnicity, we introduce MixCuBe, a cross-cultural bias benchmark, and study elements from five countries and four ethnicities. Our findings reveal that MLLMs achieve both higher accuracy and lower sensitivity to such perturbation for high-resource cultures, but not for low-resource cultures. GPT-4o, the best-performing model overall, shows up to 58% difference in accuracy between the original and perturbed cultural settings in low-resource cultures. Our dataset is publicly available at: https://huggingface.co/datasets/kyawyethu/MixCuBe.

cs.CL↗

arXiv2025

LLM-C3MOD: A Human-LLM Collaborative System for Cross-Cultural Hate Speech Moderation

Content moderation is a global challenge, yet major tech platforms prioritize high-resource languages, leaving low-resource languages with scarce native moderators. Since effective moderation depends on understanding contextual cues, this imbalance increases the risk of improper moderation due to non-native moderators' limited cultural understanding. Through a user study, we identify that non-native moderators struggle with interpreting culturally-specific knowledge, sentiment, and internet culture in the hate speech moderation. To assist them, we present LLM-C3MOD, a human-LLM collaborative pipeline with three steps: (1) RAG-enhanced cultural context annotations; (2) initial LLM-based moderation; and (3) targeted human moderation for cases lacking LLM consensus. Evaluated on a Korean hate speech dataset with Indonesian and German participants, our system achieves 78% accuracy (surpassing GPT-4o's 71% baseline), while reducing human workload by 83.6%. Notably, human moderators excel at nuanced contents where LLMs struggle. Our findings suggest that non-native moderators, when properly supported by LLMs, can effectively contribute to cross-cultural hate speech moderation.

cs.CL↗

arXiv2024

Cellular sheaf Laplacians on the set of simplices of symmetric simplicial set induced by hypergraph

We generalize cellular sheaf Laplacians on an ordered finite abstract simplicial complex to the set of simplices of a symmetric simplicial set. We construct a functor from the category of hypergraphs to the category of finite symmetric simplicial sets and define cellular sheaf Laplacians on the set of simplices of finite symmetric simplicial set induced by hypergraph. We provide formulas for cellular sheaf Laplacians and show that cellular sheaf Laplacian on an ordered finite abstract simplicial complex is exactly the ordered cellular sheaf Laplacian on the set of simplices induced by abstract simplicial complex.

math.AT↗

arXiv2024

Survey of Cultural Awareness in Language Models: Text and Beyond

Large-scale deployment of large language models (LLMs) in various applications, such as chatbots and virtual assistants, requires LLMs to be culturally sensitive to the user to ensure inclusivity. Culture has been widely studied in psychology and anthropology, and there has been a recent surge in research on making LLMs more culturally inclusive in LLMs that goes beyond multilinguality and builds on findings from psychology and anthropology. In this paper, we survey efforts towards incorporating cultural awareness into text-based and multimodal LLMs. We start by defining cultural awareness in LLMs, taking the definitions of culture from anthropology and psychology as a point of departure. We then examine methodologies adopted for creating cross-cultural datasets, strategies for cultural inclusion in downstream tasks, and methodologies that have been used for benchmarking cultural awareness in LLMs. Further, we discuss the ethical implications of cultural alignment, the role of Human-Computer Interaction in driving cultural inclusion in LLMs, and the role of cultural alignment in driving social science research. We finally provide pointers to future research based on our findings about gaps in the literature.

cs.CL↗

arXiv2024

Nonvanishing of $L$-function of some Hecke characters on cyclotomic fields

In this paper, we show the nonvanishing of some Hecke characters on cyclotomic fields. The main ingredient of this paper is a computation of eigenfunctions and the action of Weil representation at some primes including the primes above $2$. As an application, we show that for each isogeny factor of the Jacobian of the $p$-th Fermat curve where $2$ is a quadratic residue modulo $p$, there are infinitely many twists whose analytic rank is zero. Also, for a certain hyperelliptic curve over the $11$-th cyclotomic field whose Jacobian has complex multiplication, there are infinitely many twists whose analytic rank is zero.

math.NT↗

arXiv2024

The Robotic MAAO 0.7m Telescope System: Performance and Standard Photometric System

We introduce a 0.7m telescope system at the Miryang Arirang Astronomical Observatory (MAAO), a public observatory in Miryang, Korea. System integration and a scheduling program enable the 0.7m telescope system to operate completely robotically during nighttime, eliminating the need for human intervention. Using the 0.7m telescope system, we obtain atmospheric extinction coefficients and the zero-point magnitudes by observing standard stars. As a result, we find that atmospheric extinctions are moderate but they can sometimes increase depending on the weather conditions. The measured 5-sigma limiting magnitudes reach down to BVRI=19.4-19.6 AB mag for a point source with a total integrated time of 10 minutes under clear weather conditions, demonstrating comparable performance with other observational facilities operating under similar specifications and sky conditions. We expect that the newly established MAAO 0.7m telescope system will contribute significantly to the observational studies of astronomy. Particularly, with its capability for robotic observations, this system, although its primary duty is for public viewing, can be extensively used for the time-series observation of transients.

astro-ph.IM↗

arXiv2024

Algebraic description of complex conjugation on cohomology of a smooth projective hypersurface

We describe complex conjugation on the primitive middle-dimensional algebraic de Rham cohomology of a smooth projective hypersurface defined over a number field that admits a real embedding. We use Griffiths' description of the cohomology in terms of a Jacobian ring. The resulting description is algebraic up to transcendental factors explicitly given by certain periods.

math.AG↗

arXiv2023

Twisted de Rham complex for toric Calabi-Yau complete intersections and flat $F$-manifold structures

We describe the primitive middle-dimensional cohomology $\mathbb{H}$ of a compact simplicial toric complete intersection variety in terms of a twisted de Rham complex. Then this enables us to construct a concrete algorithm of formal flat $F$-manifold structures on $\mathbb{H}$ in the Calabi-Yau case by using the techniques of \cite{Park23}, which turn the twisted de Rham complex into a quantization dGBV (differential Gerstenhaber-Batalin-Vilkovisky) algebra and seek for an algorithmic solution to an associated \textit{weak primitive form.}

math.AG↗

arXiv2023

BF path integrals for elliptic curves and $p$-adic $L$-functions

We prove an arithmetic path integral formula for the inverse $p$-adic absolute values of the $p$-adic $L$-functions of elliptic curves over the rational numbers with good ordinary reduction at an odd prime $p$ based on the Iwasawa main conjecture and Mazur's control theorem. This is an elliptic curve analogue of \cite{CCKKPY}.

math.NT↗

arXiv2023

Imagine the Unseen World: A Benchmark for Systematic Generalization in Visual World Models

Systematic compositionality, or the ability to adapt to novel situations by creating a mental model of the world using reusable pieces of knowledge, remains a significant challenge in machine learning. While there has been considerable progress in the language domain, efforts towards systematic visual imagination, or envisioning the dynamical implications of a visual observation, are in their infancy. We introduce the Systematic Visual Imagination Benchmark (SVIB), the first benchmark designed to address this problem head-on. SVIB offers a novel framework for a minimal world modeling problem, where models are evaluated based on their ability to generate one-step image-to-image transformations under a latent world dynamics. The framework provides benefits such as the possibility to jointly optimize for systematic perception and imagination, a range of difficulty levels, and the ability to control the fraction of possible factor combinations used during training. We provide a comprehensive evaluation of various baseline models on SVIB, offering insight into the current state-of-the-art in systematic visual imagination. We hope that this benchmark will help advance visual systematic compositionality.

cs.CV↗

arXiv2023

Facing Off World Model Backbones: RNNs, Transformers, and S4

World models are a fundamental component in model-based reinforcement learning (MBRL). To perform temporally extended and consistent simulations of the future in partially observable environments, world models need to possess long-term memory. However, state-of-the-art MBRL agents, such as Dreamer, predominantly employ recurrent neural networks (RNNs) as their world model backbone, which have limited memory capacity. In this paper, we seek to explore alternative world model backbones for improving long-term memory. In particular, we investigate the effectiveness of Transformers and Structured State Space Sequence (S4) models, motivated by their remarkable ability to capture long-range dependencies in low-dimensional sequences and their complementary strengths. We propose S4WM, the first world model compatible with parallelizable SSMs including S4 and its variants. By incorporating latent variable modeling, S4WM can efficiently generate high-dimensional image sequences through latent imagination. Furthermore, we extensively compare RNN-, Transformer-, and S4-based world models across four sets of environments, which we have tailored to assess crucial memory capabilities of world models, including long-term imagination, context-dependent recall, reward prediction, and memory-based reasoning. Our findings demonstrate that S4WM outperforms Transformer-based world models in terms of long-term memory, while exhibiting greater efficiency during training and imagination. These results pave the way for the development of stronger MBRL agents.

cs.LG↗

arXiv2021

On the Jacobian of hyperelliptic curves $y^2 = x^5 + m^2$

In this paper, we study the algebraic rank and the analytic rank of the Jacobian of hyperelliptic curves $y^2 = x^5 + m^2$ for integers $m$. Namely, we first provide a condition on $m$ that gives a bound of the size of Selmer group and then we provide a condition on $m$ that makes $L$-functions non-vanishing. As a consequence, we construct a Jacobian that satisfies the rank part of the Birch--Swinnerton-Dyer conjecture.

math.NT↗

arXiv2021

Batalin-Vilkovisky formalism in the $p$-adic Dwork theory

The goal of this article is to develop BV (Batalin-Vilkovisky) formalism in the $p$-adic Dwork theory. Based on this formalism, we explicitly construct a $p$-adic dGBV algebra (differential Gerstenhaber-Batalin-Vilkovisky algebra) for a smooth projective complete intersection variety $X$ over a finite field, whose cohomology gives the $p$-adic Dwork cohomology of $X$, and its cochain endomorphism (the $p$-adic Dwork Frobenius operator) which encodes the information of the zeta function $X$. As a consequence, we give a modern deformation theoretic interpretation of Dwork's theory of the zeta function of $X$ and derive a formula for the $p$-adic Dwork Frobenius operator in terms of homotopy Lie morphisms and the Bell polynomials.

math.NT↗

arXiv2021

A basis of algebraic de Rham cohomology of complete intersections over a characteristic zero field

Let $\Bbbk$ be a field of characteristic 0. Let $X$ be a smooth complete intersection over $k$ of dimension $n-k$ in the projective space $\mathbf{P}^n_{k}$, for given positive integers $n$ and $k$. When $k=\mathbb{C}$, Terasoma (\cite{Ter90}) and Konno (\cite{Ko91}) provided an explicit representative (in terms of differential forms) of a basis for the primitive middle-dimensional algebraic de Rham cohomology $H_{dR,\operatorname{prim}}^{n-k}(X;\mathbb{C})$. Later Dimca constructed another explicit representative of a basis of $H_{dR,\operatorname{prim}}^{n-k}(X;\mathbb{C})$ in \cite{Dim95}. Moreover, he proved that his representative gives the same cohomology class as the previous representative of Terasoma and Konno. The goal of this article is to examine the above two different approaches without assuming that $k=\mathbb{C}$ and provide a similar comparison result for any field $k$. Dimca's argument depends heavily on the condition $k=\mathbb{C}$ and our idea is to find appropriate Cech-de Rham complexes and spectral sequences corresponding to those two approaches, which work without restrictions on $k$.

math.AG↗

arXiv2021

Polynomial realizations of period matrices of projective smooth complete intersections and their deformation

Let $X$ be a smooth complete intersection over $\mathbb{C}$ of dimension $n-k$ in the projective space $\mathbf{P}^n_{\mathbb{C}}$, for given positive integers $n$ and $k$. For a given integral homology cycle $[γ] \in H_{n-k}(X(\mathbb{C}),\mathbb{Z})$, the period integral is defined to be a linear map from the de Rham cohomology group to $\mathbb{C}$ given by $[ω] \mapsto \int_γω$. The goal of this article is to interpret this period integral as a linear map from the polynomial ring with $n+k+1$ variables to $\mathbb{C}$ and use this interpretation to develop a deformation theory of period integrals of $X$. The period matrix is an invariant defined by the period integrals of the \textit{rational} de Rham cohomology, which compares the \textit{rational} structures ($\mathbb{Q}$-subspace structures) of the de Rham cohomology over $\mathbb{C}$ and the singular homology with coefficient $\mathbb{C}$. As a main result, when $X'$ is another projective smooth complete intersection variety deformed from $X$, we provide an explicit formula for the period matrix of $X'$ in terms of the period matrix of $X$ and the Bell polynomials evaluated at the deformation data. Our result can be thought of as a modern deformation theoretic treatment of the period integrals based on the Maurer-Cartan equation of a dgla (differential graded Lie algebra).

math.AG↗