arXiv subjects

Woohyun Cho

Publications and source records attributed to Woohyun Cho.

7 recordsLinked to original sources

arXiv2026

Revisiting Residual Connections: Orthogonal Updates for Stable and Efficient Deep Networks

Residual connections are pivotal for deep neural networks, enabling greater depth by mitigating vanishing gradients. However, in standard residual updates, the module's output is directly added to the input stream. This can lead to updates that predominantly reinforce or modulate the existing stream direction, potentially underutilizing the module's capacity for learning entirely novel features. In this work, we introduce Orthogonal Residual Update: we decompose the module's output relative to the input stream and add only the component orthogonal to this stream. This design aims to guide modules to contribute primarily new representational directions, fostering richer feature learning while promoting more efficient training. We demonstrate that our orthogonal update strategy improves generalization accuracy and training stability across diverse architectures (ResNetV2, Vision Transformers) and datasets (CIFARs, TinyImageNet, ImageNet-1k), achieving, for instance, a +3.78 pp top-1 accuracy gain for ViT-B on ImageNet-1k.

cs.CV↗

arXiv2025

Spatial correlations of charge density wave order across the transition in 2H-NbSe2

Charge density waves (CDWs) involve coupled amplitude and phase degrees of freedom, but direct access to local amplitude correlations remains experimentally challenging. Here, we report cryogenic four-dimensional scanning transmission electron microscopy (4D-STEM) measurements of CDW ordering in a 2H-NbSe2 flake of 24 nm thickness, enabled by liquid helium-based cooling. By mapping the spatial distribution of CDW superlattice intensities at nanometer-scale resolution and analyzing their autocorrelations, we extract the temperature-dependent correlation length associated with the local amplitude of the CDW order parameter, independent of global phase coherence. Our results reveal that a finite local CDW amplitude is already established well above the transition temperature. When the system is cooled below the transition temperature down to 20 K, the correlation length extends to nearly 110 nm, and the local CDW amplitude is found to strongly anticorrelate with the local strain field.

cond-mat.str-el↗

arXiv2025

Laser-Induced Topological Toggle Switching at Room Temperature in the van der Waals Ferromagnet Fe3GaTe2

We demonstrate room-temperature nucleation and manipulation of topological spin textures in the van der Waals (vdW) ferromagnet, Fe3GaTe2, through laser pulse excitation. By leveraging laser-induced heating and subsequent cooling, we access the skyrmion/bubble state at low fields and achieve toggle switching between two topological spin textures - skyrmion/bubble and labyrinth. Micromagnetic simulations reveal that this switching behaviour arises from laser-induced heating and cooling. Our findings highlight the potential of vdW ferromagnets for room temperature laser-controlled non-volatile memory storage applications.

cond-mat.mtrl-sci↗

arXiv2025

MAVL: A Multilingual Audio-Video Lyrics Dataset for Animated Song Translation

Lyrics translation requires both accurate semantic transfer and preservation of musical rhythm, syllabic structure, and poetic style. In animated musicals, the challenge intensifies due to alignment with visual and auditory cues. We introduce Multilingual Audio-Video Lyrics Benchmark for Animated Song Translation (MAVL), the first multilingual, multimodal benchmark for singable lyrics translation. By integrating text, audio, and video, MAVL enables richer and more expressive translations than text-only approaches. Building on this, we propose Syllable-Constrained Audio-Video LLM with Chain-of-Thought SylAVL-CoT, which leverages audio-video cues and enforces syllabic constraints to produce natural-sounding lyrics. Experimental results demonstrate that SylAVL-CoT significantly outperforms text-based models in singability and contextual accuracy, emphasizing the value of multimodal, multilingual approaches for lyrics translation.

cs.CL↗

arXiv2025

Polymorphic spin ordering in a single-crystalline cobalt-doped Fe3GaTe2

A single crystalline system typically stabilizes a unique state for spin ordering below a critical temperature. Certain materials exhibit multiple magnetic states, driven by structural phase transitions under varying thermodynamic conditions. Recently, van der Waals magnets have demonstrated subtle interlayer exchange interactions, offering a promising approach to electrically control spin states without structural transformation. Here, we report the emergence of three distinct magnetic states, ferromagnetic ordering and both collinear and non-collinear antiferromagnetic orderings, in a layered single crystalline magnet, cobalt-doped Fe3GaTe2 ((Co, Fe)3GaTe2). These three magnetic phases occur without structural phase transitions, a phenomenon we designate as polymorphic spin ordering in the material. The introduction of 16% Co-doping in Fe3GaTe2 modulates the interlayer magnetic interaction, enabling multiple spin orderings within the same lattice system with three critical temperatures: a Curie temperature for a ferromagnetic state (Tc=210 K) and two Neel temperatures for the collinear (TN1=110 K) and non-collinear (TN2=30 K) antiferromagnetic states. Our findings are supported by magnetic force microscopy, first-principles calculations, and circular dichroism angular photoemission spectroscopy, which reveals varying spin ordering and changes in the topological band structure and Berry curvature at different temperatures within the single-crystalline (Co, Fe)3GaTe2.

cond-mat.mtrl-sci↗

arXiv2025

SEAL: Entangled White-box Watermarks on Low-Rank Adaptation

Recently, LoRA and its variants have become the de facto strategy for training and sharing task-specific versions of large pretrained models, thanks to their efficiency and simplicity. However, the issue of copyright protection for LoRA weights, especially through watermark-based techniques, remains underexplored. To address this gap, we propose SEAL (SEcure wAtermarking on LoRA weights), the universal whitebox watermarking for LoRA. SEAL embeds a secret, non-trainable matrix between trainable LoRA weights, serving as a passport to claim ownership. SEAL then entangles the passport with the LoRA weights through training, without extra loss for entanglement, and distributes the finetuned weights after hiding the passport. When applying SEAL, we observed no performance degradation across commonsense reasoning, textual/visual instruction tuning, and text-to-image synthesis tasks. We demonstrate that SEAL is robust against a variety of known attacks: removal, obfuscation, and ambiguity attacks.

cs.AI↗

arXiv2023

Singular Hall response from a correlated ferromagnetic flat nodal-line semimetal

Topological quantum phases have been largely understood in weakly correlated systems, which have identified various quantum phenomena such as spin Hall effect, protected transport of helical fermions, and topological superconductivity. Robust ferromagnetic order in correlated topological materials particularly attracts attention, as it can provide a versatile platform for novel quantum devices. Here, we report singular Hall response arising from a unique band structure of flat topological nodal lines in combination with electron correlation in an itinerant, van der Waals ferromagnetic semimetal, Fe3GaTe2, with a high Curie temperature of Tc=360 K. High anomalous Hall conductivity violating the conventional scaling, resistivity upturn at low temperature, and a large Sommerfeld coefficient are observed in Fe3GaTe2, which implies heavy fermion features in this ferromagnetic topological material. Our circular dichroism in angle-resolved photoemission spectroscopy and theoretical calculations support the original electronic features in the material. Thus, low-dimensional Fe3GaTe2 with electronic correlation, topology, and room-temperature ferromagnetic order appears to be a promising candidate for robust quantum devices.

cond-mat.str-el↗