arXiv subjects

Dongwon Kim

Publications and source records attributed to Dongwon Kim.

At least 19 recordsLinked to original sources

arXiv2026

EventCoT: Event-centric Video Chain-of-thought for Reasoning Temporal Localization

Reasoning temporal localization (RTL) requires a model to generate an answer that itself contains the time interval supporting it, coupling high-level reasoning with temporal grounding in a single response. To tackle this challenge, we propose the first event-centric video chain-of-thought framework, dubbed EventCoT. EventCoT first performs event-centric tokenization, converting the video into compact event tokens that enable efficient identification of question-relevant events. It then reasons within these events to generate the answer, grounding the time interval via embedding matching that aligns placeholder tokens with visual embeddings. EventCoT achieves state-of-the-art results on ActivityNet-RTL while using substantially fewer visual tokens than previous work, and attains strong zero-shot results on the grounded video question answering benchmark ReXTime. Our code will be released for research purposes.

cs.CV↗

arXiv2026

ACID: Action Consistency via Inverse Dynamics for Planning with World Models

Decision-time planning with action-conditioned world models has become a popular paradigm for embodied control. However, the standard planning cost judges a candidate solely by how close its predicted terminal state lies to the goal, leaving the realizability of the intermediate transitions unchecked -- a predicted trajectory can look convincing while the environment rollout drifts away from it. In this paper, we propose ACID, a decision-time planning framework that introduces cycle action consistency: the action inferred backward from a predicted transition by an inverse dynamics model should recover the one that was conditioned on. We fold this per-step residual into the planning cost via a scale-invariant adaptive weight. Across four action-conditioned world models and six tasks spanning rigid and deformable manipulation, articulated control, and visual navigation, ACID consistently improves planning and matches the baseline's accuracy with substantially less planning compute.

cs.RO↗

arXiv2026

One-Step Generalization Ratio Guided Optimization for Domain Generalization

Domain Generalization (DG) aims to train models that generalize to unseen target domains but often overfit to domain-specific features, known as undesired correlations. Gradient-based DG methods typically guide gradients in a dominant direction but often inadvertently reinforce spurious correlations. Recent work has employed dropout to regularize overconfident parameters, but has not explicitly adjusted gradient alignment or ensured balanced parameter updates. We propose GENIE (Generalization-ENhancing Iterative Equalizer), a novel optimizer that leverages the One-Step Generalization Ratio (OSGR) to quantify each parameter's contribution to loss reduction and assess gradient alignment. By dynamically equalizing OSGR via a preconditioning factor, GENIE prevents a small subset of parameters from dominating optimization, thereby promoting domain-invariant feature learning. Theoretically, GENIE balances convergence contribution and gradient alignment among parameters, achieving higher OSGR while retaining SGD's convergence rate. Empirically, it outperforms existing optimizers and enhances performance when integrated with various DG and single-DG methods.

cs.LG↗

arXiv2026

Federated Learning for Feature Generalization with Convex Constraints

Federated learning (FL) often struggles with generalization due to heterogeneous client data. Local models are prone to overfitting their local data distributions, and even transferable features can be distorted during aggregation. To address these challenges, we propose FedCONST, an approach that adaptively modulates update magnitudes based on the parameter strength of the global model. This prevents over-emphasizing well-learned parameters while reinforcing underdeveloped ones. Specifically, FedCONST employs linear convex constraints to ensure training stability and preserve locally learned generalization capabilities during aggregation. A Gradient Signal to Noise Ratio (GSNR) analysis further validates the effectiveness of FedCONST in enhancing feature transferability and robustness. As a result, FedCONST effectively aligns local and global objectives, mitigating overfitting and promoting stronger generalization across diverse FL environments, achieving state-of-the-art performance.

cs.LG↗

arXiv2026

Personalized Federated Learning for Gradient Alignment

Personalized federated learning (pFL) aims to adapt models to client specific data distributions, yet it often fails to reliably preserve personalized information. Local training is hindered by high variance gradients induced by limited and heterogeneous client data, while aggregation further distorts client specific optimization directions. To address these challenges, we propose pFLAlign, a gradient alignment framework to maintain client specific information during both local training and aggregation. pFLAlign consists of two complementary mechanisms: one adapts local gradient directions to reduce variance during client side optimization, and the other mitigates aggregation induced distortion by realigning the global model with each client's personalized direction. Theoretically, we derive pFLAlign from a PAC Bayesian analysis, which reveals how personalized gradient alignment preserves client specific information. Our experiments and ablation studies show that pFLAlign consistently improves personalization performance and training stability, achieving state of the art results.

cs.LG↗

arXiv2026

Structured State-Space Regularization for Generation-Friendly Image Tokenization

Image tokenizers play a central role in modern generative models, where the structure of the latent space critically determines the downstream generation performance. A key but underexplored property of effective latent representations is spectral organization, the ability to encode information across frequency components. In this work, we introduce structured state-space regularization, a principled approach to inducing spectral structure in latent spaces. We derive a regularization objective by revisiting state-space models (SSMs) as systems mimicking a basis function's behavior. This perspective reveals that hidden states of SSMs are induced to capture the frequency components, resulting in a novel regularizer that enforces the latent space to capture spectral structure of images. Experiments demonstrate that our regularizer improves the generative performance of image tokenizers while incurring only minimal loss in their reconstruction fidelity.

cs.CV↗

arXiv2026

Raon-Speech Technical Report

We present Raon-Speech, a top-performing 9B-parameter speech language model (SpeechLM) for English and Korean speech understanding, answering, and generation, and Raon-SpeechChat, a high-performing full-duplex extension for natural real-time conversation. Raon-Speech successfully transforms a pre-trained LLM into a SpeechLM that both understands and generates speech while preserving strong text capabilities. It trains on 1.38M hours of highly curated English and Korean speech and text datasets with the following training stages: (1) speech modules alignment, (2) end-to-end SpeechLM pre-training with knowledge distillation, and (3) multi-task preference optimization-based post-training. Across 42 English and Korean speech and text benchmarks, Raon-Speech establishes the strongest overall profile on speech-centric tasks in our comparison against eight similarly sized recent audio foundation models, including Qwen2.5-Omni and Fun-Audio-Chat, while preserving strong text question answering performance. Building upon it, Raon-SpeechChat enables natural full-duplex conversation by continual training on 119K hours of time-aligned real and synthetic dialogue data. It proceeds through three complementary training stages: (1) causal encoder adaptation, (2) full-duplex pre-training, (3) full-duplex fine-tuning for voice and role-control. On multiple full-duplex benchmarks, Raon-SpeechChat shows its clearest strengths on the turn-taking and interruption-sensitive behaviors covered by FDB v1.0, and remains competitive across the broader full-duplex evaluation suite. We open-source all model checkpoints, the training and inference pipeline, and an interactive demo.

cs.CL↗

arXiv2026

MIMO Channel Prediction via Deep Learning-based Conformal Bayes Filter

Channel prediction has emerged as an effective solution for acquiring accurate channel state information (CSI) in the presense of channel aging. Existing methods have inherent limitations, with conventional Kalman filter (KF)-based approach being vulnerable to model mismatch and deep learning (DL)-based approaches producing overconfident predictions. To address these issues, we propose a DL-based conformal Bayes filter (DCBF) that integrates DL-based prediction, conformal quantile regression (CQR), and Bayesian filtering. The proposed framework enables principled fusion of calibrated priors and observations, yielding reliable channel predictions with the calibrated uncertainty. Simulation results demonstrate that DCBF significantly improves DL-based prediction and outperforms the KF-based method.

eess.SP↗

arXiv2026

Planning in 8 Tokens: A Compact Discrete Tokenizer for Latent World Model

World models provide a powerful framework for simulating environment dynamics conditioned on actions or instructions, enabling downstream tasks such as action planning or policy learning. Recent approaches leverage world models as learned simulators, but its application to decision-time planning remains computationally prohibitive for real-time control. A key bottleneck lies in latent representations: conventional tokenizers encode each observation into hundreds of tokens, making planning both slow and resource-intensive. To address this, we propose CompACT, a discrete tokenizer that compresses each observation into as few as 8 tokens, drastically reducing computational cost while preserving essential information for planning. An action-conditioned world model that occupies CompACT tokenizer achieves competitive planning performance with orders-of-magnitude faster planning, offering a practical step toward real-world deployment of world models.

cs.CV↗

arXiv2025

Automating High Energy Physics Data Analysis with LLM-Powered Agents

We present a proof-of-principle study demonstrating the use of large language model (LLM) agents to automate a representative high energy physics (HEP) analysis. Using the Higgs boson diphoton cross-section measurement as a case study with ATLAS Open Data, we design a hybrid system that combines an LLM-based supervisor-coder agent with the Snakemake workflow manager. In this architecture, the workflow manager enforces reproducibility and determinism, while the agent autonomously generates, executes, and iteratively corrects analysis code in response to user instructions. We define quantitative evaluation metrics including success rate, error distribution, costs per specific task, and average number of API calls, to assess agent performance across multi-stage workflows. To characterize variability across architectures, we benchmark a representative selection of state-of-the-art LLMs spanning the Gemini and GPT-5 series, the Claude family, and leading open-weight models. While the workflow manager ensures deterministic execution of all analysis steps, the final outputs still show stochastic variation. Although we set the temperature to zero, other sampling parameters (e.g., top-p, top-k) remained at their defaults, and some reasoning-oriented models internally adjust these settings. Consequently, the models do not produce fully deterministic results. This study establishes the first LLM-agent-driven automated data-analysis framework in HEP, enabling systematic benchmarking of model capabilities, stability, and limitations in real-world scientific computing environments. The baseline code used in this work is available at https://huggingface.co/HWresearch/LLM4HEP. This work was accepted as a poster at the Machine Learning and the Physical Sciences (ML4PS) workshop at NeurIPS 2025. The initial submission was made on August 30, 2025.

physics.data-an↗

arXiv2025

Multi-Rate Task-Oriented Communication for Multi-Edge Cooperative Inference

The integration of artificial intelligence (AI) with the Internet of Things (IoT) enables task-oriented communication for multi-edge cooperative inference system, where edge devices transmit extracted features of local sensory data to an edge server to perform AI-driven tasks. However, the privacy concerns and limited communication bandwidth pose fundamental challenges, since simultaneous transmission of extracted features with a single fixed compression ratio from all devices leads to severe inefficiency in communication resource utilization. To address this challenge, we propose a framework that dynamically adjusts the code rate in feature extraction based on its importance to the downstream inference task by adopting a rate-adaptive quantization (RAQ) scheme. Furthermore, to select the code rate for each edge device under limited bandwidth constraint, a dynamic programming (DP) approach is leveraged to allocate the code rate across discrete code rate options. Experiments on multi-view datasets demonstrate that the proposed frameworks significantly outperform the frameworks using fixed-rate quantization, achieving a favorable balance between communication efficiency and inference performance under limited bandwidth conditions.

eess.SP↗

arXiv2025

Latent-Space Mean-Field Theory for Deep BitNet-like Training: Constrained Gradient Flows with Smooth Quantization and STE Limits

This work develops a mean-field analysis for the asymptotic behavior of deep BitNet-like architectures as smooth quantization parameters approach zero. We establish that empirical measures of latent weights converge weakly to solutions of constrained continuity equations under vanishing quantization smoothing. Our main theoretical contribution demonstrates that the natural exponential decay in smooth quantization cancels out apparent singularities, yielding uniform bounds on mean-field dynamics independent of smoothing parameters. Under standard regularity assumptions, we prove convergence to a well-defined limit that provides the mathematical foundation for gradient-based training of quantized neural networks through distributional analysis.

math.OC↗

arXiv2025

Structure and Dynamics of the Young Massive Star Cluster Westerlund 1

We present a structural analysis of the young massive star cluster Westerlund 1 (Wd 1). With multi-epoch Hubble Space Telescope (HST) observations, we measure the proper motions of $10346$ stars and determine their kinematic memberships by fitting a Gaussian mixture model to their proper motions. After correcting for extinction and completeness, we model the stellar density distribution and confirm the presence of an elongation with an eccentricity of $0.71$. The eccentricity decreases slightly with increasing mass. We fit the radial profile with the Elson, Fall, and Freeman model, observing a decrease in the core radius with increasing mass, indicative of weak but detectable mass segregation. This finding is further supported by a measured mass segregation ratio of $\Lambda_\mathrm{\rm MSR}=1.11\pm0.11$, only above $1$ by $1\sigma$, and slightly shorter minimum spanning tree length for higher mass bins. The cluster has a 1D velocity dispersion of $3.42 \pm 0.10~\mathrm{km}\,\mathrm{s}^{-1}$, suggesting it is subvirial. The subvirial state implies either exceptionally high star formation efficiency or inefficient stellar feedback caused by local gas expulsion before stars reach the cluster. The crossing time is $0.30$ Myr and the relaxation time is $0.26$ Gyr. Given the age of Wd 1 of $10.7$ Myr, we expect evident mass segregation for stars more massive than $10~M_\odot$, which accounts for the minor mass segregation found in the mass range of $1.00\unicode{x2013}12.14~M_\odot$ in this work. This suggests the overall mass segregation in Wd 1 is not primordial.

astro-ph.GA↗

arXiv2025

Democratizing Text-to-Image Masked Generative Models with Compact Text-Aware One-Dimensional Tokens

Image tokenizers form the foundation of modern text-to-image generative models but are notoriously difficult to train. Furthermore, most existing text-to-image models rely on large-scale, high-quality private datasets, making them challenging to replicate. In this work, we introduce Text-Aware Transformer-based 1-Dimensional Tokenizer (TA-TiTok), an efficient and powerful image tokenizer that can utilize either discrete or continuous 1-dimensional tokens. TA-TiTok uniquely integrates textual information during the tokenizer decoding stage (i.e., de-tokenization), accelerating convergence and enhancing performance. TA-TiTok also benefits from a simplified, yet effective, one-stage training process, eliminating the need for the complex two-stage distillation used in previous 1-dimensional tokenizers. This design allows for seamless scalability to large datasets. Building on this, we introduce a family of text-to-image Masked Generative Models (MaskGen), trained exclusively on open data while achieving comparable performance to models trained on private data. We aim to release both the efficient, strong TA-TiTok tokenizers and the open-data, open-weight MaskGen models to promote broader access and democratize the field of text-to-image masked generative models.

cs.CV↗

arXiv2024

Improving Text-based Person Search via Part-level Cross-modal Correspondence

Text-based person search is the task of finding person images that are the most relevant to the natural language text description given as query. The main challenge of this task is a large gap between the target images and text queries, which makes it difficult to establish correspondence and distinguish subtle differences across people. To address this challenge, we introduce an efficient encoder-decoder model that extracts coarse-to-fine embedding vectors which are semantically aligned across the two modalities without supervision for the alignment. There is another challenge of learning to capture fine-grained information with only person IDs as supervision, where similar body parts of different individuals are considered different due to the lack of part-level supervision. To tackle this, we propose a novel ranking loss, dubbed commonality-based margin ranking loss, which quantifies the degree of commonality of each body part and reflects it during the learning of fine-grained body part details. As a consequence, it enables our method to achieve the best records on three public benchmarks.

cs.CV↗

arXiv2024

1.58-bit FLUX

We present 1.58-bit FLUX, the first successful approach to quantizing the state-of-the-art text-to-image generation model, FLUX.1-dev, using 1.58-bit weights (i.e., values in {-1, 0, +1}) while maintaining comparable performance for generating 1024 x 1024 images. Notably, our quantization method operates without access to image data, relying solely on self-supervision from the FLUX.1-dev model. Additionally, we develop a custom kernel optimized for 1.58-bit operations, achieving a 7.7x reduction in model storage, a 5.1x reduction in inference memory, and improved inference latency. Extensive evaluations on the GenEval and T2I Compbench benchmarks demonstrate the effectiveness of 1.58-bit FLUX in maintaining generation quality while significantly enhancing computational efficiency.

cs.CV↗

arXiv2024

ATLAS EFT Results in the Top Quark Sector

With no direct evidence for new physics at the TeV scale, deviations from the Standard Model (SM) can be explored systematically through Effective Field Theories (EFTs) such as the Standard Model EFT (SMEFT). SMEFT extends the SM by introducing higher-dimensional operators parametrized by Wilson coefficients, offering a framework to probe beyond the SM (BSM) effects. This contribution highlights three recent analyses using the ATLAS Run-2 dataset at a center-of-mass energy of $\sqrt{s}$ = 13 TeV with an integrated luminosity of 140 $\text{fb}^{-1}$. Combined measurements improve the sensitivity to Wilson coefficients by reducing degeneracies and tightening constraints, exploring the potential of SMEFT in the top quark sector.

hep-ex↗

arXiv2024

Robust Bayesian Optimization via Localized Online Conformal Prediction

Bayesian optimization (BO) is a sequential approach for optimizing black-box objective functions using zeroth-order noisy observations. In BO, Gaussian processes (GPs) are employed as probabilistic surrogate models to estimate the objective function based on past observations, guiding the selection of future queries to maximize utility. However, the performance of BO heavily relies on the quality of these probabilistic estimates, which can deteriorate significantly under model misspecification. To address this issue, we introduce localized online conformal prediction-based Bayesian optimization (LOCBO), a BO algorithm that calibrates the GP model through localized online conformal prediction (CP). LOCBO corrects the GP likelihood based on predictive sets produced by LOCBO, and the corrected GP likelihood is then denoised to obtain a calibrated posterior distribution on the objective function. The likelihood calibration step leverages an input-dependent calibration threshold to tailor coverage guarantees to different regions of the input space. Under minimal noise assumptions, we provide theoretical performance guarantees for LOCBO's iterates that hold for the unobserved objective function. These theoretical findings are validated through experiments on synthetic and real-world optimization tasks, demonstrating that LOCBO consistently outperforms state-of-the-art BO algorithms in the presence of model misspecification.

cs.LG↗