arXiv subjects

Yeji Park

Publications and source records attributed to Yeji Park.

10 recordsLinked to original sources

arXiv2026

ExPerT: Personalizing LLM Responses to Users' Domain Expertise via Query-Wise Semantic and Keystroke Behavioral Cues

Large language models (LLMs) are increasingly used by end users, yet existing personalization methods relying on static profiles or text-only signals fail to capture query-specific expertise variation. We present ExPerT, a query-wise personalization framework that adapts LLM responses to users' query domain expertise by combining semantic and behavioral cues. ExPerT consists of two key components: (i) a semantic-behavioral expertise inference module that jointly interprets query text and keystroke dynamics via in-context LLM prompting, and (ii) an expertise-conditioned response generation that adapts the level of detail, terminology, and conceptual complexity. Our user study with 40 participants and 1270 queries demonstrated that ExPerT reduced expertise inference error by 65.7% compared to the strongest baseline (MAE = 0.398 vs. 1.162) and improved response satisfaction by 17.52% (from 3.71 to 4.36) on a 5-point Likert scale.

cs.HC↗

arXiv2026

Enhancing Multi-Image Understanding through Delimiter Token Scaling

Large Vision-Language Models (LVLMs) achieve strong performance on single-image tasks, but their performance declines when multiple images are provided as input. One major reason is the cross-image information leakage, where the model struggles to distinguish information across different images. Existing LVLMs already employ delimiter tokens to mark the start and end of each image, yet our analysis reveals that these tokens fail to effectively block cross-image information leakage. To enhance their effectiveness, we propose a method that scales the hidden states of delimiter tokens. This enhances the model's ability to preserve image-specific information by reinforcing intra-image interaction and limiting undesired cross-image interactions. Consequently, the model is better able to distinguish between images and reason over them more accurately. Experiments show performance gains on multi-image benchmarks such as Mantis, MuirBench, MIRB, and QBench2. We further evaluate our method on text-only tasks that require clear distinction. The method improves performance on multi-document and multi-table understanding benchmarks, including TQABench, MultiNews, and WCEP-10. Notably, our method requires no additional training or inference cost.

cs.CV↗

arXiv2025

Gravitational Wave Signatures from Lepton Number Breaking Phase Transitions with Flat Potentials

Extensions of the Standard Model typically contain ``flaton fields" defined as fields with large vacuum expectation values and almost flat potentials where scalar self-coupling is small or vanishes at tree level. Such potentials have been used to drive a secondary inflationary epoch after a primary phase of inflation, in what are called thermal inflation models. Although the primordial, high-scale inflationary epoch can solve the horizon and flatness problems, it does not always resolve difficulties associated with late-time relics produced in extensions of the Standard Model. These relics typically decay too late, injecting entropy and energetic particles that spoil successful predictions like Big Bang Nucleosynthesis. It is here that thermal inflation plays a crucial role: diluting unwanted relics by many orders of magnitude without erasing the baryon asymmetry or the large-scale structure set up by the earlier phase of inflation. The preferred scale for this phenomenon is in the range $10^6-10^8$ GeV if one considers supergravity, but without it, any scale above the EW scale is valid. We investigate a typical form of these potentials and determine what are the conditions for the potentials to develop a barrier such that when the flatons settle to the true minimum, the associated Gravitational Waves can be observed, focusing on first-order phase transitions from spontaneous lepton number breaking.

hep-ph↗

arXiv2025

Mitigating Cross-Image Information Leakage in Multi-Image Understanding with Large Vision-Language Models

Large Vision-Language Models (LVLMs) exhibit strong performance on single-image tasks. However, their performance degrades significantly when handling multi-image inputs. While this degradation has been observed in prior work, its nature remains poorly understood. We empirically observe visual elements from different images become entangled in the model's representations and responses. We refer to this phenomenon as cross-image information leakage. To address this issue, we propose FOCUS, a training-free and architecture-agnostic method. FOCUS masks all but one image with random noise, guiding the model to focus on the single clean image. This process is applied across the target images to obtain logits under partially masked contexts. These logits are aggregated and then refined using a noise-only reference input, which suppresses the leakage and yields more accurate outputs. FOCUS consistently improves performance on diverse multi-image benchmarks. We further show that FOCUS generalizes to video understanding, extending its applicability beyond static multi-image inputs. This demonstrates that FOCUS offers a general solution for enhancing multi-image reasoning without additional training or architectural modifications.

cs.CV↗

arXiv2024

Gravitational waves from a first-order phase transition of the inflaton

We explore the production of gravitational waves (GW) resulting from a first-order phase transition (FOPT) in a non-minimally coupled `Dark Higgs Inflation' model. Utilizing a dark sector scalar field as the inflaton, we demonstrate how inflationary dynamics naturally set the stage for observable FOPT. These transitions, influenced by thermal and quantum effects, generate GW spectra potentially detectable by observatories such as LISA, DECIGO, the Cosmic Explorer and the Einstein Telescope. Our study highlights the inflaton's dual role in cosmic inflation and early Universe phase transitions, presenting a unified framework to probe physics beyond the Standard Model through gravitational wave astronomy.

hep-ph↗

arXiv2024

ConVis: Contrastive Decoding with Hallucination Visualization for Mitigating Hallucinations in Multimodal Large Language Models

Hallucinations in Multimodal Large Language Models (MLLMs) where generated responses fail to accurately reflect the given image pose a significant challenge to their reliability. To address this, we introduce ConVis, a novel training-free contrastive decoding method. ConVis leverages a text-to-image (T2I) generation model to semantically reconstruct the given image from hallucinated captions. By comparing the contrasting probability distributions produced by the original and reconstructed images, ConVis enables MLLMs to capture visual contrastive signals that penalize hallucination generation. Notably, this method operates purely within the decoding process, eliminating the need for additional data or model updates. Our extensive experiments on five popular benchmarks demonstrate that ConVis effectively reduces hallucinations across various MLLMs, highlighting its potential to enhance model reliability.

cs.CV↗

arXiv2022

DeeLeMa: Missing information search with Deep Learning for Mass estimation

We introduce DeeLeMa, a deep learning-based network for the analysis of energy and momentum in high-energy particle collisions. This novel approach is specifically designed to address the challenge of analyzing collision events with multiple invisible particles, which are prevalent in many high-energy physics experiments. DeeLeMa is constructed based on the kinematic constraints and symmetry of the event topologies. We show that DeeLeMa can robustly estimate mass distribution even in the presence of combinatorial uncertainties and detector smearing effects. The approach is flexible and can be applied to various event topologies by leveraging the relevant kinematic symmetries. This work opens up exciting opportunities for the analysis of high-energy particle collision data, and we believe that DeeLeMa has the potential to become a valuable tool for the high-energy physics community.

hep-ph↗

arXiv2022

Primordial Black Holes as a Factory of Axions: Extragalactic Photons from Axions

Primordial black holes (PBHs) are significant sources of axions and axion-like particles (ALPs), provided their Hawking temperature exceeds the particles' masses. Given the predominant decay of axions into photons, the enhanced photon spectrum they generate can be feasibly detected using sensitive detectors. This paper introduces a novel methodology that elucidates the decay process for particles to traverse and decay over cosmological timescales. Specifically, we derive estimations for the photon spectrum and flux, assuming a monochromatic mass spectrum and isotropic distribution for PBHs. Encouragingly, forthcoming detectors like e-ASTROGAM are well positioned to capture this signal.

hep-ph↗

arXiv2020

Transaction-level Model Simulator for Communication-Limited Accelerators

Rapid design space exploration in early design stage is critical to algorithm-architecture co-design for accelerators. In this work, a pre-RTL cycle-accurate accelerator simulator based on SystemC transaction-level modeling (TLM), AccTLMSim, is proposed for convolutional neural network (CNN) accelerators. The accelerator simulator keeps track of each bus transaction between accelerator and DRAM, taking into account the communication bandwidth. The simulation results are validated against the implementation results on the Xilinx Zynq. Using the proposed simulator, it is shown that the communication bandwidth is severely affected by DRAM latency and bus protocol overhead. In addition, the loop tiling is optimized to maximize the performance under the constraint of on-chip SRAM size. Furthermore, a new performance estimation model is proposed to speed up the design space exploration. Thanks to the proposed simulator and performance estimation model, it is possible to explore a design space of millions of architectural options within a few tens of minutes.

cs.AR↗

arXiv2020

Light gauge boson interpretation for $(g-2)_μ$ and the $K_L \rightarrow π^0 + \text{(invisible)}$ anomaly at the J-PARC KOTO experiment

We discuss a list of possible light gauge boson interpretations for the long-standing experimental anomaly in $(g-2)_μ$ and also recent anomalous excess in $K_L \rightarrow π^0 + \text{(invisible)}$ events at the J-PARC KOTO experiment. We consider two models: $i$) $L_μ- L_τ$ gauge boson with heavy vector-like quarks and $ii$) $(L_μ- L_τ) + ε(B_3 - L_τ)$ gauge boson in the presence of right-handed neutrinos. When the light gauge boson has mass close to the neutral pion in order to satisfy the Grossman-Nir bound, the models successfully explain the anomalies simultaneously while satisfying all known experimental constraints. We extensively provide the future prospect of suggested models.

hep-ph↗