Search arXiv⌕ Search

arXiv subjects

Yiran Xu

Publications and source records attributed to Yiran Xu.

At least 19 recordsLinked to original sources
arXiv2026

Tensor Derivatives, Unified Tensor-Form Differential Equations, and Model Reduction via Partial Tucker Decomposition

This paper develops a unified tensor calculus for matrix-valued functions and their derivatives, and leverages this framework to construct efficient model reduction techniques for high-dimensional tensor differential equations. We first establish a systematic theory of tensor differentiation, wherein the derivative of a matrix with respect to another matrix is represented as a fourth-order tensor. Building on this calculus, we recast linear ordinary differential equations (ODEs) and partial differential equations(PDEs) into a compact tensor-matrix form $\frac{dX}{dt} = \A\ast X$. The general solution is expressed as $X = \exp(t\A)\ast C$, extending the matrix exponential to the tensor setting. Conditions under which the solution admits this exponential form are characterized in terms of the commutativity of the associated matrix slices. We introduce the partial Tucker decomposition (parTuckerD) to address the computational challenges posed by high-order tensor systems. On a synthetic electronic health record (EHR) tensor, parTuckerD achieves a relative reconstruction error of $0.0992$ with a $136.3\times$ compression ratio, matching the accuracy of the full TuckerD while preserving patient-level similarity structure. The results demonstrate that the proposed tensor calculus and parTuckerD framework provide a principle and computationally efficient approach for analyzing and solving high-dimensional tensor differential equations arising in data-intensive applications.

math.CA↗
arXiv2026

Trifocal Tensors in Three-View Geometry

Matrices and tensors are ubiquitous throughout computer vision. The trifocal tensor $\caT$ is a $3\times 3\times 3$ tensor that plays a vital role in three-view geometry. However, this tensor, though displayed in a third-order form, is manipulated and operated primarily in a matrix style in standard literature. This article establishes the conformal tensor algebra by replacing ad-hoc matrix collections with well-defined multilinear operators. We treat all three views symmetrically by covariant subscript indexing ($T_{ijk}$). To make precise what is new beyond notation: the classical point, line, and mixed correspondence constraints are here re-derived, but \emph{uniformly}, each as a single contraction in one shared algebra; the results that are genuinely new include the exact rank characterization of the trifocal tensor --- all mode-$k$ unfoldings $\caT[k]$ in $\RR^{3 \times 9}$ satisfy $\rank(\caT[k])=3$ for non-degenerate setups, and rank deficiency of any unfolding certifies degeneracy of the camera configuration, which yields a near-zero-cost degeneracy alarm and a validation criterion for estimated tensors; the outer-product representation $\caT = A^{\top}\times \bfb_{4} - B^{\top}\times_{2} \bfa_{4}$; and direct epipole extraction via double contractive traces. The coordinate-free multilinear operators seamlessly map to tensor auto-differentiation frameworks (e.g., PyTorch, TensorFlow); a concrete PyTorch experiment demonstrates that the induced $24$-parameter bilinear parametrization acts as a structural prior that eliminates the accuracy drift exhibited by an unstructured $27$-entry autodiff refinement.

math.AG↗
arXiv2026

When Variance Is Not an Error Map: Calibrated Uncertainty for Radiative Gaussian Splatting in Sparse-View CT

Does an uncertainty map identify where a reconstruction is wrong? In sparse-view computed tomography (CT), we find a sharp gap between whole-volume evaluation and error localization inside the object. We derive clamp-aware analytic moments for factorized Gaussian-density distributions, with a variance pass through existing rendering interfaces that is $7.9\times$ faster than a 16-sample estimator. On a 15-scene benchmark, median variance--error Spearman correlation falls from $0.846$ over the whole volume to $0.108$ in foreground. The pattern recurs across representations and acquisition settings. Two analyses help explain the discrepancy: region contrast dominates global covariance, while $72$--$96\%$ of in-object squared error is shared across independently trained members. Spread is unchanged by a common error, although shared error alone does not determine ranking. Correcting the offset between the deployed reconstruction and predictive mean improves foreground correlation by only $0.0014$. The diagnosis separates two remedies. A log-normal control improves scale transfer without restoring localization; a supervised error predictor raises foreground correlation to $0.427$ on the benchmark and $0.566$ on eight held-out human subjects under simulated acquisition. A verified retrospective re-execution on eight additional subjects retains a median of $0.537$, with the same frozen predictors. The central lesson is to validate uncertainty in the region, and against the error target, for which it will be used.

cs.CV↗
arXiv2026

FLARE: Diffusion for Hybrid Language Model

Autoregressive (AR) large language models (LLMs) have achieved broad practical success, but sequential decoding remains a key bottleneck for low-latency deployment. Recent efficient-inference work has progressed along two axes: reducing the cost of each model invocation through efficient architectures, and reducing serial decoding steps through parallel generation. Hybrid attention backbones address the former, while diffusion language models (dLLMs) pursue the latter via iterative parallel denoising. Combining these advantages remains challenging: AR-to-dLLM conversion often fails to preserve seed-checkpoint capability, and hybrid-attention recurrent states and masking constraints make diffusion training and serving nontrivial. We present FLARE, a systematic conversion framework for hybrid-attention LLMs. Our analysis identifies transfer data quality as the primary determinant of capability preservation, outweighing loss formulation and attention-mask design. The resulting framework combines a token-equal AR-and-diffusion objective, hardware-aware kernels, and unified inference, enabling one checkpoint to support both AR-style verified decoding and diffusion-style parallel denoising. Starting from strong AR checkpoints with limited post-training data, FLARE is competitive with leading open-source dLLMs across model scales and delivers consistent throughput gains over open-source dLLM baselines in single-GPU concurrent serving. Our results further suggest that practical dLLMs are limited not only by decoding algorithms, but also by transfer data quality and the training inefficiency of current block-diffusion objectives, motivating joint design of data, objectives, architectures, and inference systems.

cs.LG↗
arXiv2026

Chimera: Designing and Chinchilla-Scaling Hybrid Visual Diffusion Transformers

Visual generation increasingly requires high-resolution images, long videos, and multimodal context, making the quadratic cost of full attention prohibitive. We introduce Chimera, a hybrid visual diffusion backbone with a principled scaling recipe. Chimera processes text, image, and video tokens in one raster-ordered stream without positional embeddings. It combines Kimi Delta Attention (KDA) for long-context state tracking with O(N) complexity, interleaved Multi-head Latent Attention (MLA) for direct global interaction, and modality-aware short convolutions for local spatiotemporal context. Sparse Mixture-of-Experts (MoE) layers expand capacity while controlling activated compute. To scale this heterogeneous architecture, we introduce HeteroP, a module-wise scheme that transfers hyperparameters across width and depth according to each tensor's functional fan-in and model depth. HeteroP yields a consistently tuned family used to fit Chinchilla-style compute-optimal laws for activated model size, training-token count, and image-video data ratio. Guided by these laws, we train an 11B-parameter Chimera with 2B activated parameters. Experiments show three results. First, measured by pretraining diffusion loss, the dense backbone is 1.7x as compute-efficient as a matched full-attention Wan-2.1 2B baseline, while the complete system reaches 7.3x. Second, without length-specific fine-tuning, Chimera extrapolates zero-shot from 5-second training clips to 30-second videos, with only 6.5% FID degradation in the last five seconds. Third, the fitted laws show that compute-optimal image pretraining divides compute nearly evenly between activated model size and training-token count, whereas video pretraining modestly favors model size at higher budgets. These results establish a foundation for designing and scaling efficient long-context diffusion architectures.

cs.CV↗
arXiv2026

Smaller Models are Natural Explorers for Policy-Level Diversity in GRPO

We identify a new dimension for enhancing rollout diversity in Group Relative Policy Optimization (GRPO) for LLMs. While GRPO relies on diverse rollouts, prevailing strategies primarily increase diversity by injecting more token-level randomness, which may introduce step-wise noise and lead to incoherent trajectories. We uncover that smaller models within the same model family inherently exhibit higher policy-level diversity, indicated by their superior pass@k relative to larger counterparts as sample counts increase. Unlike token-level noise, this diversity is temporally correlated, preserves logical consistency, and provides structured exploration signals for gradient estimation. We thus propose S2L-PO (Small-to-Large Policy Optimization), a framework that leverages fixed small models as natural explorers to train larger models. To balance exploration and exploitation, we design a progressive annealing strategy that transitions from offline small-model rollouts to the large learner's own sampling. This shift elegantly avoids mid-training performance drops caused by the small model's capacity limits, achieving faster convergence and unlocking a higher performance ceiling. S2L-PO improves accuracy on diverse mathematical reasoning benchmarks (e.g., +8.8% on AIME 24 using a 1.7B explorer to guide the 8B model) while reducing rollout compute.

cs.LG↗
arXiv2026

Self-Supervised AI-Generated Image Detection: A Camera Metadata Perspective

The proliferation of AI-generated imagery poses escalating challenges for multimedia forensics, yet many existing detectors depend on assumptions about the internals of specific generative models, limiting their cross-model applicability. We introduce a self-supervised approach for detecting AI-generated images that leverages camera metadata -- specifically exchangeable image file format (EXIF) tags -- to learn features intrinsic to digital photography. Our pretext task trains a feature extractor solely on camera-captured photographs by classifying categorical EXIF tags (\emph{e.g.}, camera model and scene type) and pairwise-ranking ordinal and continuous EXIF tags (\emph{e.g.}, focal length and aperture value). Using these EXIF-induced features, we first perform one-class detection by modeling the distribution of photographic images with a Gaussian mixture model and flagging low-likelihood samples as AI-generated. We then extend to binary detection that treats the learned extractor as a strong regularizer for a classifier of the same architecture, operating on high-frequency residuals from spatially scrambled patches. Extensive experiments across various generative models demonstrate that our EXIF-induced detectors substantially advance the state of the art, delivering strong generalization to in-the-wild samples and robustness to common benign image perturbations. The code and model are publicly available at https://github.com/Ekko-zn/SDAIE.

cs.CV↗
arXiv2026

Color Matters: Demosaicing-Guided Color Correlation Training for Generalizable AI-Generated Image Detection

As realistic AI-generated images threaten digital authenticity, we address the generalization failure of generative artifact-based detectors by exploiting the intrinsic properties of the camera imaging pipeline. Concretely, we investigate color correlations induced by the color filter array (CFA) and demosaicing, and propose a Demosaicing-guided Color Correlation Training (DCCT) framework for AI-generated image detection. By simulating the CFA sampling pattern, we decompose each color image into a single-channel input (as the condition) and the remaining two channels as the ground-truth targets (for prediction). A self-supervised U-Net is trained to model the conditional distribution of the missing channels from the given one, parameterized via a mixture of logistic functions. Our theoretical analysis reveals that DCCT targets a provable distributional difference in color-correlation features between photographic and AI-generated images. By leveraging these distinct features to construct a binary classifier, DCCT achieves state-of-the-art generalization and robustness, significantly outperforming prior methods across over 20 unseen generators.

cs.CV↗
arXiv2026

SIN-Bench: Tracing Native Evidence Chains in Long-Context Multimodal Scientific Interleaved Literature

Evaluating whether multimodal large language models truly understand long-form scientific papers remains challenging: answer-only metrics and synthetic "Needle-In-A-Haystack" tests often reward answer matching without requiring a causal, evidence-linked reasoning trace in the document. We propose the "Fish-in-the-Ocean" (FITO) paradigm, which requires models to construct explicit cross-modal evidence chains within native scientific documents. To operationalize FITO, we build SIN-Data, a scientific interleaved corpus that preserves the native interleaving of text and figures. On top of it, we construct SIN-Bench with four progressive tasks covering evidence discovery (SIN-Find), hypothesis verification (SIN-Verify), grounded QA (SIN-QA), and evidence-anchored synthesis (SIN-Summary). We further introduce "No Evidence, No Score", scoring predictions when grounded to verifiable anchors and diagnosing evidence quality via matching, relevance, and logic. Experiments on eight MLLMs show that grounding is the primary bottleneck: Gemini-3-pro achieves the best average overall score (0.573), while GPT-5 attains the highest SIN-QA answer accuracy (0.767) but underperforms on evidence-aligned overall scores, exposing a gap between correctness and traceable support.

cs.CL↗
arXiv2025

DuetSVG: Unified Multimodal SVG Generation with Internal Visual Guidance

Recent vision-language model (VLM)-based approaches have achieved impressive results on SVG generation. However, because they generate only text and lack visual signals during decoding, they often struggle with complex semantics and fail to produce visually appealing or geometrically coherent SVGs. We introduce DuetSVG, a unified multimodal model that jointly generates image tokens and corresponding SVG tokens in an end-to-end manner. DuetSVG is trained on both image and SVG datasets. At inference, we apply a novel test-time scaling strategy that leverages the model's native visual predictions as guidance to improve SVG decoding quality. Extensive experiments show that our method outperforms existing methods, producing visually faithful, semantically aligned, and syntactically clean SVGs across a wide range of applications.

cs.CV↗
arXiv2025

Rethinking Prompt Design for Inference-time Scaling in Text-to-Visual Generation

Achieving precise alignment between user intent and generated visuals remains a central challenge in text-to-visual generation, as a single attempt often fails to produce the desired output. To handle this, prior approaches mainly scale the visual generation process (e.g., increasing sampling steps or seeds), but this quickly leads to a quality plateau. This limitation arises because the prompt, crucial for guiding generation, is kept fixed. To address this, we propose Prompt Redesign for Inference-time Scaling, coined PRIS, a framework that adaptively revises the prompt during inference in response to the scaled visual generations. The core idea of PRIS is to review the generated visuals, identify recurring failure patterns across visuals, and redesign the prompt accordingly before regenerating the visuals with the revised prompt. To provide precise alignment feedback for prompt revision, we introduce a new verifier, element-level factual correction, which evaluates the alignment between prompt attributes and generated visuals at a fine-grained level, achieving more accurate and interpretable assessments than holistic measures. Extensive experiments on both text-to-image and text-to-video benchmarks demonstrate the effectiveness of our approach, including a 15% gain on VBench 2.0. These results highlight that jointly scaling prompts and visuals is key to fully leveraging scaling laws at inference-time. Visualizations are available at the website: https://subin-kim-cv.github.io/PRIS.

cs.CV↗
arXiv2025

RELIC: Interactive Video World Model with Long-Horizon Memory

A truly interactive world model requires three key ingredients: real-time long-horizon streaming, consistent spatial memory, and precise user control. However, most existing approaches address only one of these aspects in isolation, as achieving all three simultaneously is highly challenging-for example, long-term memory mechanisms often degrade real-time performance. In this work, we present RELIC, a unified framework that tackles these three challenges altogether. Given a single image and a text description, RELIC enables memory-aware, long-duration exploration of arbitrary scenes in real time. Built upon recent autoregressive video-diffusion distillation techniques, our model represents long-horizon memory using highly compressed historical latent tokens encoded with both relative actions and absolute camera poses within the KV cache. This compact, camera-aware memory structure supports implicit 3D-consistent content retrieval and enforces long-term coherence with minimal computational overhead. In parallel, we fine-tune a bidirectional teacher video model to generate sequences beyond its original 5-second training horizon, and transform it into a causal student generator using a new memory-efficient self-forcing paradigm that enables full-context distillation over long-duration teacher as well as long student self-rollouts. Implemented as a 14B-parameter model and trained on a curated Unreal Engine-rendered dataset, RELIC achieves real-time generation at 16 FPS while demonstrating more accurate action following, more stable long-horizon streaming, and more robust spatial-memory retrieval compared with prior work. These capabilities establish RELIC as a strong foundation for the next generation of interactive world modeling.

cs.CV↗
arXiv2025

PAD3R: Pose-Aware Dynamic 3D Reconstruction from Casual Videos

We present PAD3R, a method for reconstructing deformable 3D objects from casually captured, unposed monocular videos. Unlike existing approaches, PAD3R handles long video sequences featuring substantial object deformation, large-scale camera movement, and limited view coverage that typically challenge conventional systems. At its core, our approach trains a personalized, object-centric pose estimator, supervised by a pre-trained image-to-3D model. This guides the optimization of deformable 3D Gaussian representation. The optimization is further regularized by long-term 2D point tracking over the entire input video. By combining generative priors and differentiable rendering, PAD3R reconstructs high-fidelity, articulated 3D representations of objects in a category-agnostic way. Extensive qualitative and quantitative results show that PAD3R is robust and generalizes well across challenging scenarios, highlighting its potential for dynamic scene understanding and 3D content creation.

cs.CV↗
arXiv2025

HALO: Human-Aligned End-to-end Image Retargeting with Layered Transformations

Image retargeting aims to change the aspect-ratio of an image while maintaining its content and structure with less visual artifacts. Existing methods still generate many artifacts or fail to maintain original content or structure. To address this, we introduce HALO, an end-to-end trainable solution for image retargeting. Since humans are more sensitive to distortions in salient areas than non-salient areas of an image, HALO decomposes the input image into salient/non-salient layers and applies different wrapping fields to different layers. To further minimize the structure distortion in the output images, we propose perceptual structure similarity loss which measures the structure similarity between input and output images and aligns with human perception. Both quantitative results and a user study on the RetargetMe dataset show that HALO achieves SOTA. Especially, our method achieves an 18.4% higher user preference compared to the baselines on average.

cs.CV↗
arXiv2025

Cropper: Vision-Language Model for Image Cropping through In-Context Learning

The goal of image cropping is to identify visually appealing crops in an image. Conventional methods are trained on specific datasets and fail to adapt to new requirements. Recent breakthroughs in large vision-language models (VLMs) enable visual in-context learning without explicit training. However, downstream tasks with VLMs remain under explored. In this paper, we propose an effective approach to leverage VLMs for image cropping. First, we propose an efficient prompt retrieval mechanism for image cropping to automate the selection of in-context examples. Second, we introduce an iterative refinement strategy to iteratively enhance the predicted crops. The proposed framework, we refer to as Cropper, is applicable to a wide range of cropping tasks, including free-form cropping, subject-aware cropping, and aspect ratio-aware cropping. Extensive experiments demonstrate that Cropper significantly outperforms state-of-the-art methods across several benchmarks.

cs.CV↗
arXiv2025

Global smooth solutions to Navier-Stokes equations with large initial data in critical space

In this paper, we investigate the existence of a unique global smooth solution to the three-dimensional incompressible Navier-Stokes equations and provide a concise proof. We establish a new global well-posedness result that allows the initial data to be arbitrarily large within the critical space $\dot{B}^{-1}_{\infty,\infty}$, while still satisfying the nonlinear smallness condition.

math.AP↗
arXiv2024

Optimizing Quantum Communication for Quantum Data Centers with Reconfigurable Networks

Distributed Quantum Computing (DQC) enables scalability by interconnecting multiple QPUs. Among various DQC implementations, quantum data centers (QDCs), which utilize reconfigurable optical switch networks to link QPUs across different racks, are becoming feasible in the near term. However, the latency of cross-rack communications and dynamic reconfigurations poses unique challenges to quantum communication, significantly increasing the overall latency and exacerbating qubit decoherence. In this paper, we introduce a new optimization space to parallelize cross-rack communications and avoid frequent reconfigurations, which incurs additional in-rack communications that can be further minimized. Based on this, we propose a flexible scheduler that improves communication efficiency while preventing deadlocks and congestion caused by the flexibility. Through a comprehensive evaluation, we show that our approach reduces the overall latency by a factor of 8.02, thereby mitigating qubit decoherence, with a small overhead.

quant-ph↗