arXiv subjects

Yujin Kang

Publications and source records attributed to Yujin Kang.

7 recordsLinked to original sources

arXiv2026

A.X K2 Technical Report

We introduce A.X K2, a 688B-parameter Mixture-of-Experts (MoE) language model trained from scratch as a high-performance foundation for \emph{agentic} applications. Trained on approximately 8.5T tokens---fewer than its predecessor, A.X K1---on a smaller but higher-quality mixture with substantially expanded agentic and software-engineering data, it nonetheless improves over A.X K1 across the board, by over 30 percentage points on some benchmarks, reflecting large gains in token efficiency. To support long contexts efficiently, we introduce Sparse Gated Attention (SGA), which combines sparse attention with gated attention, and adopt Gated Norm (GN) to stabilize large-scale training. SGA is trained natively at 128K through a \emph{sparse} indexer warmup that optimizes the indexer against its own sparse top-$k$ selection rather than the dense attention distribution, making adaptation markedly cheaper: each query reads only 2,048 positions, yet long-context quality is unchanged and A.X K2 scores 94.6 on RULER out to 256K. The outlier suppression of GN in turn keeps 4-bit NVFP4 serving within one point of FP8 accuracy. A simple yet effective Think-Fusion recipe further lets users switch between thinking and non-thinking modes within a single unified model. Extensive evaluations show that A.X K2 performs competitively against strong open-weight baselines, matching or exceeding them on math and Korean-language benchmarks.

cs.AI↗

arXiv2026

Low Spatial Cost CCZ Magic State Factory

We propose a design framework for reconstructing gate-based magic state distillation protocols as compact joint-measurement architectures implementable with the surface code. The goal is to reduce the surface-code resource cost of a magic state factory while preserving the logical function and error-detection structure of the distillation protocol. We construct a reduced architecture for implementing an eight-to-three CCZ distillation protocol using smaller surface-code patches. The proposed factory preserves the single-fault-detection property and the leading-order error suppression of the protocol, while producing CCZ magic states with lower spatial cost than the design of Gidney and Fowler. The proposed design perspective can also be applied to T-state factories and other multiqubit non-Clifford resource-state factories. Our approach provides a framework for extending the design space of surface-code magic state factories beyond a single CCZ layout optimization.

quant-ph↗

arXiv2026

Defect-Adaptive Lattice Surgery on Irregular Boundary Surface-Code Patches

Defect-adaptive surface-code methods have substantially advanced the construction of valid logical patches on imperfect hardware, but fault-tolerant computation also requires executable logical oper ations on the resulting irregular geometries. We formulate the seam-boundary defect problem: how to perform a lattice-surgery merge when the intended seam intersects deformed boundaries, disabled checks, and gauge-inferred super-stabilizers. We introduce a defect-adaptive lattice-surgery method that reconstructs the target joint logical parity from the seam-related measurements available on the irregular merged patch, together with constraints inherited from the separated pre-merge code space. The reconstruction is expressed as a compact GF(2) binary-support synthesis problem. If the requested parity is realizable, the solution gives an executable parity-extraction rule over raw, schedule-tagged gauge outcomes; otherwise, it certifies a parity-synthesis failure rather than conflat ing it with patch invalidity. The framework accommodates boundary data-qubit defects, seam-check ancilla defects, and gauge-inferred seam super-checks within a single synthesis layer. Circuit-level samples of the synthesized merge operation show improved compile yield, preserved effective dis tance, and only modest success-conditioned logical-error overhead relative to the defect-free merge reference; an explicit ZZ-merge sampling check confirms the expected transposed-geometry behav ior under the same success-conditioned observable construction. More broadly, the results identify certified parity synthesis as a compilation layer between defect-adaptive patch construction and executable fault-tolerant logical operations on imperfect surface-code hardware.

quant-ph↗

arXiv2025

ReBrain: Brain MRI Reconstruction from Sparse CT Slice via Retrieval-Augmented Diffusion

Magnetic Resonance Imaging (MRI) plays a crucial role in brain disease diagnosis, but it is not always feasible for certain patients due to physical or clinical constraints. Recent studies attempt to synthesize MRI from Computed Tomography (CT) scans; however, low-dose protocols often result in highly sparse CT volumes with poor through-plane resolution, making accurate reconstruction of the full brain MRI volume particularly challenging. To address this, we propose ReBrain, a retrieval-augmented diffusion framework for brain MRI reconstruction. Given any 3D CT scan with limited slices, we first employ a Brownian Bridge Diffusion Model (BBDM) to synthesize MRI slices along the 2D dimension. Simultaneously, we retrieve structurally and pathologically similar CT slices from a comprehensive prior database via a fine-tuned retrieval model. These retrieved slices are used as references, incorporated through a ControlNet branch to guide the generation of intermediate MRI slices and ensure structural continuity. We further account for rare retrieval failures when the database lacks suitable references and apply spherical linear interpolation to provide supplementary guidance. Extensive experiments on SynthRAD2023 and BraTS demonstrate that ReBrain achieves state-of-the-art performance in cross-modal reconstruction under sparse conditions.

cs.CV↗

arXiv2025

Optimizing Cross-Domain Transfer for Universal Machine Learning Interatomic Potentials

Accurate yet transferable machine-learning interatomic potentials (MLIPs) are essential for accelerating materials and chemical discovery. However, most universal MLIPs overfit to narrow datasets or computational protocols, limiting their reliability across chemical and functional domains. We introduce a transferable multi-domain training strategy that jointly optimizes universal and task-specific parameters through selective regularization, coupled with a domain-bridging set (DBS) that aligns potential-energy surfaces across datasets. Systematic ablation experiments show that small DBS fractions (0.1%) and targeted regularization synergistically enhance out-of-distribution generalization while preserving in-domain fidelity. Trained on fifteen open databases spanning molecules, crystals, and surfaces, our model, SevenNet-Omni, achieves state-of-the-art cross-domain accuracy, including adsorption-energy errors below 0.06 eV on metallic surfaces and 0.1 eV on metal-organic frameworks. Despite containing only 0.5% r$^2$SCAN data, SevenNet-Omni reproduces high-fidelity r$^2$SCAN energetics, demonstrating effective cross-functional transfer from large PBE datasets. This framework offers a scalable route toward universal, transferable MLIPs that bridge quantum-mechanical fidelities and chemical domains.

cond-mat.mtrl-sci↗

arXiv2024

MIMO Detection under Hardware Impairments: Data Augmentation With Boosting

This paper addresses a data detection problem for multiple-input multiple-output (MIMO) communication systems with hardware impairments. To facilitate maximum likelihood (ML) data detection without knowledge of nonlinear and unknown hardware impairments, we develop novel likelihood function (LF) estimation methods based on data augmentation and boosting. The core idea of our methods is to generate multiple augmented datasets by injecting noise with various distributions into seed data consisting of online received signals. We then estimate the LF using each augmented dataset based on either the expectation maximization (EM) algorithm or the kernel density estimation (KDE) method. Inspired by boosting, we further refine the estimated LF by linearly combining the multiple LF estimates obtained from the augmented datasets. To determine the weights for this linear combination, we develop three methods that take different approaches to measure the reliability of the estimated LFs. Simulation results demonstrate that both the EM- and KDE-based LF estimation methods offer significant performance gains over existing LF estimation methods. Our results also show that the effectiveness of the proposed methods improves as the size of the augmented data increases.

eess.SP↗

arXiv2024

Vector Quantization for Deep-Learning-Based CSI Feedback in Massive MIMO Systems

This paper presents a finite-rate deep-learning (DL)-based channel state information (CSI) feedback method for massive multiple-input multiple-output (MIMO) systems. The presented method provides a finite-bit representation of the latent vector based on a vector-quantized variational autoencoder (VQ-VAE) framework while reducing its computational complexity based on shape-gain vector quantization. In this method, the magnitude of the latent vector is quantized using a non-uniform scalar codebook with a proper transformation function, while the direction of the latent vector is quantized using a trainable Grassmannian codebook. A multi-rate codebook design strategy is also developed by introducing a codeword selection rule for a nested codebook along with the design of a loss function. Simulation results demonstrate that the proposed method reduces the computational complexity associated with VQ-VAE while improving CSI reconstruction performance under a given feedback overhead.

eess.SP↗