arXiv subjects

Seong Jae Hwang

Publications and source records attributed to Seong Jae Hwang.

At least 19 recordsLinked to original sources

arXiv2026

Controlling Motion Transfer in Diffusion Transformers via Attention Heads

Diffusion Transformers (DiTs) have advanced video generation with high-quality, temporally coherent results. However, extending them to motion transfer, which requires following reference motion while aligning with a target prompt, remains challenging due to limited understanding of motion and structure representations within DiTs. We analyze video DiTs at the attention-head level and identify distinct heads specialized for motion and spatial structure. Based on this insight, we propose a head-aware controllable motion transfer framework that requires no parameter updates. Our method refines motion cues from motion-specialized heads via semantic correspondence guidance and preserves structure through selective feature injection. This head-level control not only enables accurate motion transfer but also provides an interpretable foundation for controllable video generation with DiTs.

cs.CV↗

arXiv2026

QWERTY: Training-Free Motion Control via Query-Warped Video Diffusion Transformers

Video diffusion transformers (DiTs) generate high-fidelity and temporally coherent videos, yet motion control remains implicit, primarily relying on text prompts. As a result, achieving desired motion often requires extensive prompt engineering and repeated resampling. While fine-tuning models with additional spatial prompts (e.g., bounding boxes or point trajectories) enables explicit control, it demands substantial data curation and computation, and may compromise the generative capabilities of pretrained models. Consequently, training-free motion control using such spatial prompts has been explored in U-Net-based video diffusion models, but remains largely unexplored for DiTs. We introduce QWERTY, a training-free framework that enables flexible motion control in pretrained image-to-video DiTs via user-defined object warping and optical flow. We carefully manipulate the 3D full attention of DiTs by warping the frame-invariant semantic subspace of queries. We find that the noise predicted by the query-warped DiT naturally guides the diffusion trajectory toward the desired motion, and further show that leveraging this noise as self-guidance for latent optimization improves control stability and visual quality. Experiments show that QWERTY achieves the most effective motion control among existing training-free approaches on a recent image-to-video DiT, with performance comparable to fine-tuning-based methods.

cs.CV↗

arXiv2026

Contrastive and Adaptive Multi-modal Masked Autoencoder for Spatial Transcriptomics

The high cost of spatial transcriptomics (ST) has driven extensive studies into predicting gene expression directly from H&E histology images. However, this prediction task faces an inherent limitation, as tissue morphology alone provides insufficient information to fully resolve underlying gene expression. To address this limitation, a recent study leverages partial gene expression to guide the prediction process alongside histology images. Building on this paradigm, we approach the prediction task as a spatial imputation problem, employing a Masked Autoencoder (MAE) to utilize a small fraction of gene expression as genetic anchors for inferring whole-slide gene expression profiles. Specifically, we propose a bio-saliency score and a learning-to-rank strategy to adaptively identify the most informative spots within the tissue. Based on these identified spots, our framework selects contiguous regions as genetic anchors to ensure suitability for real-world ST profiling hardware. To effectively leverage these anchors, we design a cross-modal joint encoder that integrates visual and genetic modalities. By aligning the selected anchors with their corresponding visual features via contrastive learning, the encoder generates robust joint representations to accurately predict gene expression across the whole slide. Notably, our framework consistently surpasses existing methods in both histology-only prediction and spatial imputation, achieving superior accuracy even without genetic anchors and further excelling with as little as 10% transcriptomic coverage. Our code is available at https://github.com/Kyyle2114/CAMMST.

cs.CV↗

arXiv2026

Anatomically Consistent TMJ Disc Segmentation via Semantic Anchoring and Clinical Priors

Segmenting the temporomandibular joint (TMJ) disc from MRI is essential for accurate diagnosis of internal derangement, yet it remains unreliable in practice due to its small size, low contrast, and morphological variability. Existing methods, primarily adapted from general segmentation architectures, often produce fragmented or anatomically inconsistent masks, leading to unstable measurements of disc position and shape for downstream diagnosis. To address these challenges, we propose TISC, a TMJ disc segmentation framework that integrates semantic anchoring with clinical metadata-guided boundary refinement. The framework first establishes robust disc localization in the foundation model feature space via a Prototypical Semantic Anchoring (PSA) module that aggregates adjacent-slice MedDINOv3 features and derives a prototype-driven similarity map. It then performs targeted boundary refinement through a Clinical-Metadata Point Refinement (C-MPR) module, with point-wise predictions modulated by Mouth Open Limitation (MOL), a clinical indicator associated with disc displacement without reduction. On a large-scale cohort of 2,488 PD MRI volumes from 1,300 patients, our method achieves up to a 4.96 Dice improvement over strong baselines across diverse architectures, delivering more anatomically coherent and clinically reliable TMJ disc segmentation.

eess.IV↗

arXiv2026

Physics in 2-Steps: Locking Motion Priors Before Visual Refinement Erases Them

Image-to-Video diffusion models leverage input images to generate visually stunning content, yet frequently produce motion that violates physical laws. We reveal a surprising finding: a 2-step generation often exhibits better physical consistency than a 50-step output from the same model. Through spectral analysis, we trace this to phase erosion during denoising; the phase degrades significantly (dropping by $\approx 18\%$ from step 2 to step 50), whereas the magnitude remains relatively stable. Building on this insight, we propose PhaseLock, a training-free framework that preserves the valid motion priors from few-step inference throughout the denoising trajectory. Rather than relying on full-step inference for physical consistency, PhaseLock extracts a motion prior from just 2 steps and enforces it onto high-fidelity generation via Latent Delta Guidance. Our approach effectively mitigates phase degradation, improving physical consistency by an average of 6.2 points across diverse models while largely maintaining visual fidelity, with negligible overhead ($1.06\times$ time, $1.02\times$ memory) and reduced reliance on expensive external guidance methods ($\sim5\times$ time). Project Page: https://dnwjddl.github.io/phaselock

cs.CV↗

arXiv2026

How Noisy Poses Break Inverse Dynamics: Analysis and Mitigation for Video-Based Joint Torque Estimation

Recent advances in monocular 3D human pose estimation enable accurate body tracking from video. However, translating these kinematic estimates into physical quantities, such as joint torques, remains challenging due to noise amplification through inverse dynamics. In this work, we provide a systematic analysis of how pose estimation noise propagates through the inverse dynamics pipeline. We present three key findings: (1) pose noise is amplified by approximately 1,000x when computing joint torques via numerical differentiation, (2) proximal joints (spine, hips) are up to 10x more sensitive to noise than distal joints (wrists, hands), and (3) low-pass filtering before differentiation substantially reduces this amplification. To enable this analysis, we develop SMPL-Dynamics, a fully differentiable inverse dynamics module for the SMPL body model that requires no external physics simulators. Our module supports end-to-end gradient computation, and we demonstrate this through differentiable pose refinement, which reduces torque error by 93% with negligible change in pose.

cs.CV↗

arXiv2026

Mitigating Mask Prior Drift and Positional Attention Collapse in Large Diffusion Vision-Language Models

Large diffusion vision-language models (LDVLMs) have recently emerged as a promising alternative to autoregressive models, enabling parallel decoding for efficient inference and leveraging bidirectional attention for global context. Despite these advances, their behavior under long-form generation remains underexplored. In this work, we show that existing LDVLMs suffer from repetitive generation and degraded visual grounding, and identify two underlying causes. First, repetitive generation originates from a mask token prior: since generation tokens are initialized as mask tokens, their hidden representations progressively drift toward a shared prior direction over generation steps. Second, a fundamental misalignment between the positional attention bias and the iterative unmasking process suppresses attention toward informative visual tokens, degrading visual grounding. Based on these insights, we propose a training-free approach, introducing Mask Prior Suppression and Monotonic RoPE Scaling to mitigate mask prior drift and positional attention collapse during decoding. Experiments on general multimodal benchmarks and visual grounding tasks demonstrate improvements over baseline LDVLMs, with robust gains on long-form description benchmarks. Our results show that these failures can be effectively addressed with a lightweight, plug-and-play strategy that requires no additional training and generalizes across diverse LDVLM architectures.

cs.CV↗

arXiv2026

Towards Continuous Sign Language Conversation from Isolated Signs

Sign language is the primary language for many Deaf and Hard-of-Hearing (DHH) signers, yet most conversational AI systems still mediate interaction through spoken or written language. This spoken-language-centered interface can limit access for signers for whom spoken or written language is not the most accessible medium, motivating direct sign-to-sign conversational modeling. However, sentence-level sign video data are expensive to collect and annotate, leaving existing sign translation and production models with limited vocabulary coverage and weak open-domain generalization. We address this bottleneck by constructing continuous sign conversations from isolated signs: large-scale labeled isolated clips are collected as lexically grounded motion primitives and recomposed into sign-language-ordered utterances derived from existing dialogue corpora. We introduce SignaVox-W, which provides, to our knowledge, the largest labeled isolated-sign vocabulary to date, and SignaVox-U, a continuous 3D sign conversation dataset built from SignaVox-W. To bridge structural mismatch between spoken and signed languages, we use a retrieval-guided spoken-to-gloss translator; to bridge independently collected isolated clips, we propose BRAID, a diffusion Transformer that performs duration alignment and co-articulatory boundary inpainting. With the resulting data, we train SignaVox, a direct sign-to-sign conversational model that generates 3D body, hand, and facial motion responses from prior signing context without spoken-language text or externally provided glosses at inference time. Quantitative and qualitative evaluations show improved isolated-to-continuous motion quality, stronger response-level semantic alignment, and scalable signer-centered interaction that better supports visual-spatial articulation.

cs.CV↗

arXiv2026

Rethinking Graph Convolution for 2D-to-3D Hand Pose Lifting

Graph convolutional networks (GCNs) are widely used for 3D hand pose estimation, where the hand skeleton is encoded as a fixed adjacency graph. We revisit whether this is the most effective way to incorporate hand topology in 2D-to-3D lifting. In this paper, we perform controlled, parameter-matched ablations on the FPHA benchmark and show that standard multi-head self-attention consistently outperforms GCN baselines. Even when the GCN is strengthened with multi-hop adjacency and matched parameter count, self-attention reduces MPJPE from 12.36 mm to 10.09 mm. A skeleton-constrained graph attention network recovers most of this gap, indicating that input-dependent aggregation is a major source of improvement, while fully connected attention yields additional gains. We further show that hand topology is most effective when introduced as a soft structural prior through graph-distance positional encoding, rather than as a hard adjacency constraint. These results suggest that, for hand pose lifting, adaptive spatial attention is a more effective inductive bias than fixed graph convolution.

cs.CV↗

arXiv2026

Real-Time Visual Attribution Streaming in Thinking Model

We present an amortized framework for real-time visual attribution streaming in multimodal thinking models. When these models generate code from a screenshot or solve math problems from images, their long reasoning traces should be grounded in visual evidence. However, verifying this reliance is challenging: faithful causal methods require costly repeated backward passes or perturbations, while raw attention maps offer instant access, they lack causal validity. To resolve this, we introduce an amortized approach that learns to estimate the causal effects of semantic regions directly from the rich signals encoded in attention features. Across five diverse benchmarks and four thinking models, our approach achieves faithfulness comparable to exhaustive causal methods while enabling visual attribution streaming, where users observe grounding evidence as the model reasons, not after. Our results demonstrate that real-time, faithful attribution in multimodal thinking models is achievable through lightweight learning, not brute-force computation.

cs.CV↗

arXiv2026

Delaunay Canopy: Building Wireframe Reconstruction from Airborne LiDAR Point Clouds via Delaunay Graph

Reconstructing building wireframe from airborne LiDAR point clouds yields a compact, topology-centric representation that enables structural understanding beyond dense meshes. Yet a key limitation persists: conventional methods have failed to achieve accurate wireframe reconstruction in regions afflicted by significant noise, sparsity, or internal corners. This failure stems from the inability to establish an adaptive search space to effectively leverage the rich 3D geometry of large, sparse building point clouds. In this work, we address this challenge with Delaunay Canopy, which utilizes the Delaunay graph as a geometric prior to define a geometrically adaptive search space. Central to our approach is Delaunay Graph Scoring, which not only reconstructs the underlying geometric manifold but also yields region-wise curvature signatures to robustly guide the reconstruction. Built on this foundation, our corner and wire selection modules leverage the Delaunay-induced prior to focus on highly probable elements, thereby shaping the search space and enabling accurate prediction even in previously intractable regions. Extensive experiments on the Building3D Tallinn city and entry-level datasets demonstrate state-of-the-art wireframe reconstruction, delivering accurate predictions across diverse and complex building geometries.

cs.CV↗

arXiv2026

FEAST: Fully Connected Expressive Attention for Spatial Transcriptomics

Spatial Transcriptomics (ST) provides spatially-resolved gene expression, offering crucial insights into tissue architecture and complex diseases. However, its prohibitive cost limits widespread adoption, leading to significant attention on inferring spatial gene expression from readily available whole slide images. While graph neural networks have been proposed to model interactions between tissue regions, their reliance on pre-defined sparse graphs prevents them from considering potentially interacting spot pairs, resulting in a structural limitation in capturing complex biological relationships. To address this, we propose FEAST (Fully connected Expressive Attention for Spatial Transcriptomics), an attention-based framework that models the tissue as a fully connected graph, enabling the consideration of all pairwise interactions. To better reflect biological interactions, we introduce negative-aware attention, which models both excitatory and inhibitory interactions, capturing essential negative relationships that standard attention often overlooks. Furthermore, to mitigate the information loss from truncated or ignored context in standard spot image extraction, we introduce an off-grid sampling strategy that gathers additional images from intermediate regions, allowing the model to capture a richer morphological context. Experiments on public ST datasets show that FEAST surpasses state-of-the-art methods in gene expression prediction while providing biologically plausible attention maps that clarify positive and negative interactions. Our code is available at https://github.com/starforTJ/ FEAST.

cs.CV↗

arXiv2026

ViKey: Enhancing Temporal Understanding in Videos via Visual Prompting

Recent advancements in Video Large Language Models (VideoLLMs) have enabled strong performance across diverse multimodal video tasks. To reduce the high computational cost of processing dense video frames, efficiency-oriented methods such as frame selection have been widely adopted. While effective at minimizing redundancy, these methods often cause notable performance drops on tasks requiring temporal reasoning. Unlike humans, who can infer event progression from sparse visual cues, VideoLLMs frequently misinterpret temporal relations when intermediate frames are omitted. To address this limitation, we explore visual prompting (VP) as a lightweight yet effective way to enhance temporal understanding in VideoLLMs. Our analysis reveals that simply annotating each frame with explicit ordinal information helps the model perceive temporal continuity. This visual cue also supports frame-level referencing and mitigates positional ambiguity within a sparsely sampled sequence. Building on these insights, we introduce ViKey, a training-free framework that combines VP with a lightweight Keyword-Frame Mapping (KFM) module. KFM leverages frame indices as dictionary-like keys to link textual cues to the most relevant frames, providing explicit temporal anchors during inference. Despite its simplicity, our approach substantially improves temporal reasoning and, on some datasets, preserves dense-frame baseline performance with as few as 20% of frames.

cs.CV↗

arXiv2026

Anchoring and Rescaling Attention for Semantically Coherent Inbetweening

Generative inbetweening (GI) seeks to synthesize realistic intermediate frames between the first and last keyframes beyond mere interpolation. As sequences become sparser and motions larger, previous GI models struggle with inconsistent frames with unstable pacing and semantic misalignment. Since GI involves fixed endpoints and numerous plausible paths, this task requires additional guidance gained from the keyframes and text to specify the intended path. Thus, we give semantic and temporal guidance from the keyframes and text onto each intermediate frame through Keyframe-anchored Attention Bias. We also better enforce frame consistency with Rescaled Temporal RoPE, which allows self-attention to attend to keyframes more faithfully. TGI-Bench, the first benchmark specifically designed for text-conditioned GI evaluation, enables challenge-targeted evaluation to analyze GI models. Without additional training, our method achieves state-of-the-art frame consistency, semantic fidelity, and pace stability for both short and long sequences across diverse challenges.

cs.CV↗

arXiv2026

Interpretable Motion-Attentive Maps: Spatio-Temporally Localizing Concepts in Video Diffusion Transformers

Video Diffusion Transformers (DiTs) have been synthesizing high-quality video with high fidelity from given text descriptions involving motion. However, understanding how Video DiTs convert motion words into video remains insufficient. Furthermore, while prior studies on interpretable saliency maps primarily target objects, motion-related behavior in Video DiTs remains largely unexplored. In this paper, we investigate concrete motion features that specify when and which object moves for a given motion concept. First, to spatially localize, we introduce GramCol, which adaptively produces per-frame saliency maps for any text concept, including both motion and non-motion. Second, we propose a motion-feature selection algorithm to obtain an Interpretable Motion-Attentive Map (IMAP) that localizes motion spatially and temporally. Our method discovers concept saliency maps without the need for any gradient calculation or parameter update. Experimentally, our method shows outstanding localization capability on the motion localization task and zero-shot video semantic segmentation, providing interpretable and clearer saliency maps for both motion and non-motion concepts.

cs.CV↗

arXiv2025

Rare Text Semantics Were Always There in Your Diffusion Transformer

Starting from flow- and diffusion-based transformers, Multi-modal Diffusion Transformers (MM-DiTs) have reshaped text-to-vision generation, gaining acclaim for exceptional visual fidelity. As these models advance, users continually push the boundary with imaginative or rare prompts, which advanced models still falter in generating, since their concepts are often too scarce to leave a strong imprint during pre-training. In this paper, we propose a simple yet effective intervention that surfaces rare semantics inside MM-DiTs without additional training steps, data, denoising-time optimization, or reliance on external modules (e.g., large language models). In particular, the joint-attention mechanism intrinsic to MM-DiT sequentially updates text embeddings alongside image embeddings throughout transformer blocks. We find that by mathematically expanding representational basins around text token embeddings via variance scale-up before the joint-attention blocks, rare semantics clearly emerge in MM-DiT's outputs. Furthermore, our results generalize effectively across text-to-vision tasks, including text-to-image, text-to-video, and text-driven image editing. Our work invites generative models to reveal the semantics that users intend, once hidden yet ready to surface.

cs.AI↗

arXiv2025

Interpreting vision transformers via residual replacement model

How do vision transformers (ViTs) represent and process the world? This paper addresses this long-standing question through the first systematic analysis of 6.6K features across all layers, extracted via sparse autoencoders, and by introducing the residual replacement model, which replaces ViT computations with interpretable features in the residual stream. Our analysis reveals not only a feature evolution from low-level patterns to high-level semantics, but also how ViTs encode curves and spatial positions through specialized feature types. The residual replacement model scalably produces a faithful yet parsimonious circuit for human-scale interpretability by significantly simplifying the original computations. As a result, this framework enables intuitive understanding of ViT mechanisms. Finally, we demonstrate the utility of our framework in debiasing spurious correlations.

cs.CV↗

arXiv2025

Interpreting Attention Heads for Image-to-Text Information Flow in Large Vision-Language Models

Large Vision-Language Models (LVLMs) answer visual questions by transferring information from images to text through a series of attention heads. While this image-to-text information flow is central to visual question answering, its underlying mechanism remains difficult to interpret due to the simultaneous operation of numerous attention heads. To address this challenge, we propose head attribution, a technique inspired by component attribution methods, to identify consistent patterns among attention heads that play a key role in information transfer. Using head attribution, we investigate how LVLMs rely on specific attention heads to identify and answer questions about the main object in an image. Our analysis reveals that a distinct subset of attention heads facilitates the image-to-text information flow. Remarkably, we find that the selection of these heads is governed by the semantic content of the input image rather than its visual appearance. We further examine the flow of information at the token level and discover that (1) text information first propagates to role-related tokens and the final token before receiving image information, and (2) image information is embedded in both object-related and background tokens. Our work provides evidence that image-to-text information flow follows a structured process, and that analysis at the attention-head level offers a promising direction toward understanding the mechanisms of LVLMs.

cs.CV↗