Searcharxiv⌕ Search

arXiv subjects

Seungryong Kim

Publications and source records attributed to Seungryong Kim.

At least 19 recordsLinked to original sources

Refin3R: Geometry-Guided Refinement for Feed-Forward 3D Reconstruction

Feed-forward 3D reconstruction models predict dense point maps from multiple images in a single pass. Without explicit geometric constraints, however, their predictions are often inconsistent across views and lack fine-grained precision. Recent methods incorporate geometric priors such as camera pose or depth, but use them merely as auxiliary inputs, process each view independently, or refine the geometry only where the priors are available. We present Refin3R, which refines feed-forward point maps with Structure-from-Motion (SfM) points, propagating SfM geometry over the image grid. In its dual-branch architecture, a 2D branch extends SfM geometry beyond its coverage using geometry-aware backbone features, while a 3D branch reasons over the SfM points of all views, capturing multi-view structure that the image grid alone cannot represent. The two branches exchange information bidirectionally to jointly refine each view, yielding dense and accurate point maps. Across ScanNet++, ETH3D, Tanks\&Temples, and out-of-domain human and surgical scenes, Refin3R consistently outperforms feed-forward and geometry-guided methods in point map accuracy while running faster than the strongest geometry-guided baseline.

cs.CV↗

BirdDreams: Joint Perspective and BEV Video Generation for a Plausible, Reactive Driving World Model

Driving world models aim to simulate how a scene unfolds under different ego actions. This requires plausible motion of surrounding vehicles as well as precise ego-trajectory control. Learning vehicle motion from perspective video alone is challenging because motion is entangled with appearance and vehicles can be occluded. Prior work therefore conditions video generation on layouts that specify vehicle motion, providing control over surrounding vehicles but requiring their future trajectories as inputs. We introduce BirdDreams, an autoregressive driving world model that jointly generates perspective video and a schematic bird's-eye-view (BEV) video, making scene layout part of the predicted future. The BEV, rendered from scene annotations for training, depicts roads and vehicles as simple shapes at a fixed metric scale, providing a top-down view of their positions and motion with reduced occlusion between vehicles. Perspective video, by contrast, retains scene details that the BEV omits, such as traffic signals and object appearance. Joint attention between two video diffusion transformers, one per stream, lets each stream inform the other during generation. The predicted BEV also provides a geometric interface for ego control: we translate and rotate its latest frame along an input ego trajectory and use the resulting frames as a soft condition that shows how the scene would appear to move if only the ego vehicle moved. At inference, BirdDreams generates both videos from a single camera image without requiring an HD map or vehicle-state inputs. On PhysicalAI-AV and nuScenes, BirdDreams follows input ego trajectories more faithfully than prior driving world models and generates more plausible ego motion without ego-action conditioning. Human evaluation shows that BirdDreams outperforms a perspective-only baseline in generating realistic surrounding-vehicle behavior.

cs.CV↗

Motion Cues from Image-based Point Tracking for LiDAR Scene Flow Estimation

Self-supervised LiDAR scene flow approaches rely on pseudo-labels that distinguish static and dynamic points, providing targeted supervision. These labels are typically derived from LiDAR geometry, making them unreliable in sparse or occluded regions. The resulting noise misguides motion supervision and degrades scene flow learning, yet existing methods remain largely dependent on the same geometric evidence. To address this, we introduce TrackCue, a tracking-guided framework that transfers dense image trajectories into reliable static-dynamic supervision. TrackCue anchors tracking queries to projected LiDAR points, yielding image-space trajectories that provide motion evidence beyond sparse LiDAR geometry. Since image-space displacement also reflects ego-vehicle motion, we present a visually consistent motion compensation strategy that compares each tracked trajectory with the rigid trajectory induced by ego-motion alone. It admits motion evidence only where both trajectories are valid and the deviation persists across frames, isolating true object motion while suppressing transient tracking errors. Finally, we perform visual motion cue lifting, which associates image-space object motion cues with LiDAR points for label refinement. As a result, TrackCue substantially improves the precision of dynamic pseudo-labels, providing more reliable supervision and leading to better self-supervised scene flow estimation.

cs.CV↗

Factorized MLLM Embeddings for Composed Image Retrieval

Composed Image Retrieval (CIR) aims to retrieve a target image given a reference image and a modification text. Although recent approaches leveraging Multimodal Large Language Models (MLLMs) have shown significant progress, we argue that the MLLM's compositional understanding remains underutilized. This stems from directly mapping composed queries into a retrieval embedding via contrastive supervision alone, without explicitly resolving the intended visual modifications. In this work, we introduce Factorized MLLM Embeddings (FactEmbed), which decouples compositional query comprehension from building discriminative embeddings. Rather than directly projecting the composed query into a single embedding without internal reasoning, FactEmbed employs a set of factorized latents that simulate the intended visual modifications via an intermediate target imagination objective. This process encourages the latents to ground into target semantics before forming discriminative representations for retrieval. Subsequently, to propagate these factorized representations throughout the model for retrieval, we optimize a factor-level contrastive loss while distilling the decomposed representations into a deeper attention layer of the MLLM. Finally, for efficient inference, we retain the [EOS] token to establish a global-level embedding on top of these factorized latents, while we exploit the mean-pooled factorized latents to rerank a short candidate list at negligible cost, as both are produced by the same forward pass. Extensive evaluations demonstrate that FactEmbed outperforms prior works across CIR benchmarks.

cs.CV↗

DreamX: Video Diffusion Model is a Universal Latent Action Learner

Learning a universal action representation from video requires modeling dynamics that are consistent across different executions of the same action yet invariant to changes in embodiment and appearance. Existing latent action models infer actions from visual transitions, but their representations often entangle the underlying action with irrelevant visual variations, limiting transfer across embodiments. We introduce DreamX, which repurposes a pretrained video diffusion model as a universal latent action learner. Our approach extracts compact latent actions from the video model's rich spatiotemporal features, compressing each latent video frame by $\boldsymbol{300\times}$ ($\boldsymbol{1{,}200 \rightarrow 4}$ tokens). In addition, by leveraging the video model's generative prior, these latent actions capture action-relevant dynamics disentangled from visual appearance. The resulting latent space serves as a shared action interface across embodiments (human and robot), environments, and action parameterizations. On LARY-Bench, DreamX achieves state-of-the-art performance among latent action models. Moreover, actions encoded in one setting remain effective in another: the latent space supports any-to-any video transfer and zero-shot action transfer across diverse embodiments and environments. Furthermore, this shared interface also enables policy learning in a compact latent action space, reducing the computational cost of video-diffusion-based policies.

cs.CV↗

Trace4D: Grounded Spatiotemporal Reasoning via 4D State Sequences

Understanding how objects and their spatial relationships evolve in 3D space over time remains challenging for vision-language models (VLMs), particularly under changing viewpoints. Recent approaches incorporate geometric information into reasoning, but explicit supervision for temporally ordered 4D state evolution and its use as intermediate evidence remains limited. We introduce \textbf{Trace4D}, which trains VLMs to predict 4D state sequences and reason over their inter-timestep changes alongside temporal visual observations. Its 4D-grounded chain-of-thought (CoT) explicitly represents 4D states, analyzes temporal visual context, and connects both as evidence for the answer. We construct Trace4D-Train from in-the-wild videos with 600K 4D perception examples and approximately 70K grounded QA examples, and train with joint supervised fine-tuning followed by geometry-aware GRPO. At inference, Trace4D predicts 4D states and reasoning directly from video without reference annotations or an external geometric model. Experiments on Trace4D-Bench and VLM4D show improved QA accuracy over the base VLMs and benefits from reasoning over 4D state changes and temporal visual observations.

cs.CV↗

L3G-World: Real-Time Infinite World Generation with Latent 3D Gaussian Memory

Interactive world models generate visual content conditioned on user actions, yet maintaining spatial consistency over long-horizon exploration in real time remains challenging. Existing approaches rely on either temporal context or explicit 3D memory, both of which scale poorly because memory grows with the length of the generated stream. Moreover, explicit 3D memories condition latent video generators by re-encoding RGB renderings, which precludes real-time generation. In this paper, we introduce a Latent 3D Gaussian Memory (L3G) that combines the efficiency of feed-forward reconstruction with a compact, rapidly renderable scene representation. Building on this memory, we propose L3G-World, in which a memory updater and a video diffusion transformer (DiT) alternate autoregressively to generate spatially consistent worlds. The memory updater spawns Gaussians in newly observed regions and refines existing ones, so that the memory grows with scene extent rather than stream length. Each Gaussian further stores a VAE latent feature, allowing the memory to be rendered directly into the DiT's latent space without VAE re-encoding. Experiments show that L3G-World achieves state-of-the-art generation quality and long-horizon spatial consistency, with up to $38.6\times$ fewer 3D primitives, $3.2\times$ faster memory updates, and $359\times$ faster memory rendering relative to the most efficient prior 3D-memory baseline for each metric, enabling real-time world generation at 9 FPS.

cs.CV↗

Repurposing Video Diffusion Transformers for Cross-View Temporal Object Correspondence

Cross-view object correspondence identifies the same object across views. However, existing methods often rely on frame-wise matching or sequential tracking, which can propagate correspondence errors under appearance changes or occlusions. Therefore, we introduce cross-view temporal object correspondence, which requires preserving object identity across both views and time from a query mask. To address this, we propose XTrack, a unified framework which repurposes a pretrained video DiT with semantic and spatiotemporal priors for RGB-to-mask latent transport, jointly modeling cross-view matching and temporal tracking. However, naive video DiTs do not guarantee accurate cross-view grounding and often attend to semantically similar yet incorrect objects across views. Thus, we introduce focal cross-view attention alignment, hard negative conditioning, and local context crop strategies to ensure reliable grounding. Finally, beyond these architectural enhancements, we also tackle the lack of dense annotation by introducing a pseudo-label curation pipeline that converts sparse Ego-Exo4D annotations into dense supervision to enable effective video-level training. Our results demonstrate the effectiveness of XTrack as a unified framework, ensuring robust temporal consistency and identity preservation.

cs.CV↗

ThinkWorld: World Imagination Reinforces Chain-of-Dynamics Reasoning

Vision-Language-Action (VLA) models have made substantial progress in robotic manipulation, but their generalization remains constrained by the diversity and coverage of action-labeled robot demonstrations. Human videos provide much broader behavioral and environmental experience, while learning from them faces a fundamental trade-off between scalable data and explicit action grounding. We introduce ThinkWorld, a VLA learning framework that grounds action reasoning in world dynamics and strengthens it through reinforcement learning with world imagination. ThinkWorld represents each action thought as a Chain-of-Dynamics (CoD), a sequence of discrete motion tokens that captures the intermediate dynamics connecting the current world to a future state and serves as a common action representation across embodiments. We first initialize CoD with action-labeled demonstrations, then predict the future induced by each CoD and reinforce those whose predictions better match the observed outcome. This enables videos without action annotations to provide direct supervision from observed state transitions, allowing human and robot experience to refine the same reasoning space. Experiments in simulation and on real robots demonstrate effective human-to-robot transfer and strong generalization to unseen tasks and environments.

cs.RO↗

FL4N: Feed-forward Latent Model for 4D Novel View Synthesis

4D novel view synthesis (NVS) aims to render dynamic scenes across viewpoints and timestamps, which requires jointly inferring 3D structure and temporal dynamics. Existing monocular feed-forward methods often rely on explicit 4D Gaussian representations, which are prone to temporal inconsistency and severe rendering artifacts. We present FL4N, the first feed-forward latent 4D NVS model. Instead of relying on explicit 4D primitives, FL4N encodes a 4D scene prior into a unified latent representation across space and time. Our core insight is that learning temporal dynamics as residuals on top of a pretrained static latent prior enables effective 4D modeling with limited dynamic training data. To this end, we introduce a lightweight dynamic encoder that leverages static scene features to disentangle object motion from camera-induced parallax. Trained solely with an image synthesis objective, FL4N achieves state-of-the-art 4D NVS quality while rendering in real time. Its learned latent representation generalizes to real-world videos and disentangles viewpoint from time.

cs.CV↗

ReAnchor: Reusing Rejected Verifier States for Diffusion Drafting

Block diffusion drafters accelerate speculative decoding by generating multiple draft tokens in a single forward pass. However, each pass produces only a fixed-length block from a single target-generated anchor, limiting how far the draft can extend. Meanwhile, target verification computes hidden states even at positions beyond the first rejection, but these states are typically discarded rather than reused. We propose ReAnchor, a training-free method that reanchors a pretrained drafter at shifted positions using discarded verifier states to generate parallel draft blocks. These blocks improve predictions within the original block and extend coverage beyond its endpoint. At sampling temperature of T=0, ReAnchor achieves its largest average gains on Gemma 4-12B, improving acceptance length by 19.4% and decoding speed by 17.9% across eight benchmarks. Averaged equally over four target models and eight benchmarks, it improves acceptance length by 15.2% and decoding speed by 13.4% with pretrained DFlash drafters.

cs.CL↗

VideoMaMa++: Temporally Consistent Video Matting via Preserve-and-Refine Token

Diffusion-based video matting has recently emerged as a leading approach, leveraging video diffusion priors from internet-scale pretraining to achieve strong zero-shot generalization from limited synthetic data. However, the heavy compute of these diffusion backbones limits inference to a short chunk of just 6 to 24 frames, so even ordinary videos span multiple chunks. Because each chunk is matted independently from coarse mask guidance, existing methods fail to maintain consistency across chunk boundaries, producing flickering edges and discontinuous transparency. We address this with a mask-guided diffusion-matting framework built on two ideas. The RGB and mask conditioning are represented as separate token streams that interact through full spatio-temporal self-attention, letting the model adaptively combine fine RGB texture with coarse mask layout rather than committing to a fixed channel-wise fusion. On top of this, we propose a novel pair of learnable embeddings, the preserve token and the refine token, that act as a per-frame conditioning interface and enable a sliding-window inference scheme in which already-generated mattes are propagated across chunk boundaries for local temporal consistency, complemented by a globally shared reference anchor at every chunk to prevent error accumulation in long sequences without any additional training. Across standard video matting benchmarks and in-the-wild long footage, our framework outperforms prior diffusion-based methods in both fine-detail accuracy and temporal consistency, and uniquely maintains consistency across chunk boundaries on long videos.

cs.CV↗

Speculative Forcing for Efficient Autoregressive Video Generation

Few-step autoregressive video generators enable streaming generation, yet they run every token through the final denoising step, regardless of how much refinement it needs. We find this uniform computation is often unnecessary: many tokens already closely match their eventual full-step outputs in the initial steps, while others require refinement until the end. We introduce Speculative Forcing, a framework that treats intermediate clean estimates as drafts and predicts when individual tokens can be safely accepted early. Once accepted, a token is frozen as its final output and undergoes no further denoising, while its cached keys and values remain available as context for the remaining tokens. We explore two alternative verifier designs: a training-free one based on similarity to clean content and stability across steps, and a learned one that reads the frozen generator's intermediate features to estimate each token's remaining gap to its full-step output. Across short-video, long-video, and world generation, Speculative Forcing preserves the quality of full-step sampling while substantially reducing computation. On LongLive, it matches the four-step generator with ~1.7 denoising steps per token, running over 1.5x faster at 41.7 and 37.8 fps. On Lingbot-World 2.0, where camera motion continually changes the visible content, it maintains quality with ~2.5 denoising steps per token and runs 1.35x faster at 24.5 fps.

cs.CV↗

Can Learning to Answer Spatial Questions Improve Novel-View Generation?

Generating novel views from limited observations requires preserving both the appearance and spatial structure of a scene. Although recent approaches exploit fine-grained geometric cues, they do not explicitly model high-level semantic and spatial properties such as object identity, attributes, and inter-object relationships. In this work, we investigate whether 3D language understanding can provide complementary supervision for novel-view generation. We introduce Ask4NV, built on a pretrained unified multimodal model with a Mixture-of-Transformers (MoT) architecture. Our model incorporates camera geometry into both VAE and ViT representations and leverages understanding features during generation through joint self-attention with a modality-aware attention masking strategy. Experiments demonstrate strong image quality and geometric consistency in novel-view generation, alongside competitive 3D language understanding. Further analysis shows that joint training improves cross-view correspondence in VAE representations and similarity in a 3D-aware feature space.

cs.CV↗

PLAYGROUND: Progressive Layout Generation with Render-Grounded Planning

Generating editable graphic designs requires arranging visual and textual elements into a coherent composition while preserving their content and editability. Although recent MLLM-based approaches employ iterative generation and visual refinement to reflect the real-world design workflows, their high-level plans are predetermined prior to execution, limiting their adaptability as the design evolves. To address this limitation, we propose PlayGround, a progressive layout generation framework that dynamically adapts to the current rendered canvas. PlayGround establishes a progressive design loop comprising a Progressive Planner, a Layout Generator, and a Visual Verifier, guided by a Structure-Aware Retriever. Given an initial set of unordered elements, the Structure-Aware Retriever initially infers a coarse layout structure to retrieve structurally similar reference designs. Subsequently, at each turn, the Progressive Planner examines the current rendered canvas alongside the remaining elements, forms the next functional group, and plans its composition. The Layout Generator then predicts editable layout attributes, while the Visual Verifier evaluates rendered candidates and provides feedback when refinement is necessary. To train the Progressive Planner with step-wise supervision, we construct a two-stage data generation pipeline that extracts functional groups and turn-level planning trajectories from completed designs. Experiments on Crello and LICA demonstrate the effectiveness of PlayGround over existing methods. Our code and weights will be publicly released.

cs.CV↗

Error-Conditioned Neural Solvers

Neural surrogate models offer fast approximate mappings from PDE parameters to solutions, but they typically treat solving as a purely statistical task: once trained, they struggle to correct their own constraint violations and extrapolate beyond the training distribution. Recent hybrid methods promote physical correctness by targeting the PDE residual via gradient descent or Gauss--Newton steps, but inherit the compute cost and instability of the underlying classical optimizers. We show, theoretically and empirically, that numerically minimizing the PDE residual can be an unreliable proxy for reconstruction accuracy in ill-conditioned systems, explaining why these methods often do not make accurate predictions despite achieving low residuals. We propose error-conditioned Neural Solvers (ENS), built on a different principle: rather than an optimization target, the PDE residual field is passed as a direct input to the network at each iteration, enabling it to read the spatial structure of its own errors and learn an update policy to iteratively correct its predictions. Across four PDE families, ENS attains the highest prediction accuracy in the large majority of settings, with gains reaching $10\times$ on turbulent Kolmogorov flow, while avoiding the expensive compute cost of hybrid methods. ENS's learned correction policy generalizes under distribution shift, including zero-shot parameter changes and cross-equation transfer, where its relative advantage is largest in the ill-conditioned regimes where residual minimization is least reliable. Project website: https://neuralsolver.github.io/.

cs.LG↗

Multi-Agent Egocentric World Model with Fine-Grained Embodied Interaction

Egocentric world models predict first-person observations conditioned on an agent's actions, but most focus on a single agent. Real embodied settings often involve multiple agents that act and interact within a shared environment. Existing multi-agent world models rely on coarse actions like locomotion, camera control, or discrete commands, leaving fine-grained embodied interactions underexplored. We formulate multi-agent egocentric world modeling as synchronized ego-stream generation for multiple agents interacting through fine-grained actions in a shared world. This requires cross-view action consistency, shared-environment consistency, and consistent propagation of interaction-induced state updates. We propose Multi-agent Egocentric World Model (ME-World), which jointly denoises multiple ego streams in a shared token sequence, conditions each stream on all agents' target-view poses, and grounds generation with shared environment memory. We train and evaluate on real and synthetic multi-agent data and introduce shared-world consistency metrics for environment, update, and identity consistency. Experiments show ME-World improves shared-world consistency, action control, identity preservation, and video quality over existing methods.

cs.CV↗

GRACE: Generation-aware latent compression for efficient video generation

Highly compressed video autoencoders offer an effective way to accelerate video diffusion models, as the Diffusion Transformer (DiT) operates on far fewer tokens. However, such autoencoders are challenging to train, since a higher compression ratio degrades reconstruction quality and recovering it requires more channels, which is known to slow the convergence of the DiT. The compressed latent also differs from the one the DiT was trained on, so the pretrained DiT must be either retrained from scratch or adapted at considerable cost. Compressing the autoencoder the DiT was trained with appears to preserve compatibility, yet optimizing it for reconstruction alone still shifts the latent away from the distribution the DiT has learned. To address this, we propose Generation-Aware Latent Compression for Efficient Video Generation (GRACE), a two-stage framework that compresses a pretrained video autoencoder while keeping it compatible with the pretrained DiT. Specifically, we keep a frozen base latent from the pretrained encoder and learn a residual latent for the information lost under stronger compression, while aligning the compressed latent with the pretrained latent in the feature space of the frozen DiT so that the autoencoder is optimized for generation. We then adapt the DiT with lightweight fine-tuning and asymmetric denoising, where the base is denoised ahead of the residual. GRACE reduces the token count of Wan2.1-I2V-14B by 8x and its latency by 11.1x at 480x832x81, while matching the generation quality of the pretrained pipeline before compression on VBench.

cs.CV↗