arXiv subjects

Yuchen Lin

Publications and source records attributed to Yuchen Lin.

At least 19 recordsLinked to original sources

arXiv2026

Ring Forcing: Towards Precise Long-Term Memory for Autoregressive Video Diffusion

Scaling video generation to long durations reveals a critical bottleneck: current models lack robust long-term memory. This deficiency can be studied along two critical aspects: object permanence, the ability to precisely reproduce the appearance of objects upon re-entry; and memory capacity, the ability to process ultra-long context and use information from distant history. Robust long-term memory requires both: object permanence without sufficient context handling limits the temporal scope, while long context length without permanence fails to maintain identity. To address this, we present Ring Forcing, an autoregressive video diffusion framework designed to robustly construct and precisely utilize long-term memory. Our ring-structured training strategy enforces retrieval from distant history, effectively reconciling the trade-off between strict historical adherence and generative diversity. To expand memory capacity, we introduce a compression and timestep composition strategy. Under fixed sequence length constraints, this method extends the effective historical span to minutes-long durations and achieves a comprehensive receptive field over the entire history. Furthermore, we present a sparse RoPE mechanism to enable flexible, scalable memory adaptation while fully exploiting pre-trained priors. Extensive experiments demonstrate that Ring Forcing achieves superior minutes-long coherence and object permanence, significantly outperforming state-of-the-art methods.

cs.CV↗

arXiv2026

iVISION-2DCD: A Long-Term Change Detection Dataset for Large-Scale Outdoor Construction Monitoring

Automation in construction is essential for reducing costs and human errors in large-scale projects. We approach the construction progress monitoring from the aspect of detecting changes in construction sites. As construction buildings continue to evolve in geometry and appearance over time, change detection need to be performed from arbitrary camera viewpoints. This necessitates developing 2D Change Detection (2DCD) algorithms that operate robustly across diverse camera perspectives at construction sites. While developing and evaluating such systems is data-intensive, no open-source benchmark dataset exists at the intersection of 2D change detection and construction automation research. Data collection using Unmanned Aerial Vehicles (UAVs) is gaining its popularity in outdoor large-scale surveying. However, in active construction sites conducting drone missions equipped with high-end sensors imposes safety concerns. Flight trajectory and collected camera viewpoints can be significantly limited. To address this critical gap, we introduce iVISION-2DCD, a large-scale synthetically generated dataset from dense LiDAR point clouds with photorealistic input images and accurate ground truth annotations. Our dataset formally defines the problem of viewpoint-robust 2DCD at construction sites and captures the inherent complexities of real-world deployment. In this paper, we present our systematic methodology for synthetic data generation, developing novel view synthesis techniques to overcome bi-temporal alignment and viewpoint diversity challenges, and implementing semi-automated semantic segmentation with change label generation while preserving challenging real-world cases. Benchmark evaluations using state-of-the-art 2DCD algorithms demonstrate that iVISION-2DCD poses novel research challenges for the computer vision and robotics communities.

cs.CV↗

arXiv2026

Automatic-differentiation-enabled dynamic parameter retrieval with sub-pulse-width resolution

Time-resolved terahertz time-domain spectroscopy (THz-TDS) is a phase-sensitive tool in condensed matter physics for tracking photoinduced non-equilibrium dynamics of low-energy elementary excitations. However, the measured response function, optical conductivity $\sigma(\omega,t_{pp})$, becomes unreliable in reporting the state of matter when material properties drastically change on a timescale comparable to or less than the probe pulse duration, obscuring the sub-pulse-width dynamics. To resolve this issue, we present a full-waveform inversion framework inspired by the multi-dimensional retrieval philosophy of frequency-resolved optical gating (FROG). By leveraging the automatic differentiation (AD) technique and the two-dimensional time-domain signal $E(t_{g},t_{pp})$, we show one can uniquely solve the inverse problem, at the sub-pulse-width resolution, of retrieving physical observables that are still well-defined, i.e., time-dependent scattering rate $\gamma(t)$, plasma frequency $\omega_\mathrm{p}(t)$ and resonance frequency $\omega_0(t)$, while the response functions are not. Further optimization by gradient-based routines (Adam + L-BFGS) via JAX makes the method exceptionally robust against experimental noise and probe pulse distortions. The validity of the AD-enabled methodology is benchmarked both by a self-consistent numerical approach and by experimental data from real ultrafast THz spectroscopy measurements.

cond-mat.str-el↗

arXiv2026

Quantum Light Nano-Imaging

Entanglement and quantum correlations are central to the physics of quantum materials, yet they have remained notoriously difficult to access experimentally. Accessing these phenomena in solids requires quantum optical probes that operate at the native length and time scales of material excitations, below the diffraction limit of light. Developing the requisite tools has previously been infeasible due to the weak intensities of state-of-the-art quantum light sources and the inefficiency of light coupling in near-field light-matter interactions. In this work, we address these challenges and report the development of a quantum light scattering-type scanning near-field optical microscope (q-SNOM) that enables quantum-optical studies of solid-state systems with nanoscale spatial resolution. As a first demonstration, we visualize the self-interference of single hybrid light-matter polaritons in the prototypical van der Waals semiconductor MoS2. We also introduce a polaritonic time-of-flight metrology that exploits the temporal correlations among entangled photons to observe the quasiparticle propagation dynamics at femtosecond time scales. This work establishes a new experimental paradigm for exploring quantum effects in materials at the nanoscale.

cond-mat.mes-hall↗

arXiv2026

Signatures of quantum chaos in phonon-polariton billiards

We use scanning near-field optical microscopy to image hyperbolic phonon polaritons in hexagonal boron nitride (hBN) billiards with integrable and chaotic geometries. In Sinai billiards, we observe irregular mode patterns consistent with quantum scarring, together with an unexpected sensitivity to weak probe perturbations. These random-wave features coexist with non-chaotic one-dimensional boundary modes arising from nontrivial polariton reflection at the billiard edge. As the billiard boundary becomes increasingly complex, the Fourier transforms of the measured signals evolve toward ring-like structures consistent with Berry's random-wave conjecture. We develop a numerical framework based on the Helmholtz equation with generalized boundary conditions that encode angle-dependent reflection phase shifts. The calculated level statistics exhibit a crossover from Poisson-like behavior in integrable billiards to Wigner-Dyson-like behavior in chaotic geometries, with small deviations from the canonical form arising from nonlinear boundary conditions that require self-consistent bulk-boundary analysis. Theoretical analysis based on dissipative Green's functions qualitatively reproduces the near-field data. These results establish mesoscopic van der Waals billiards as a rich platform for studying generalized chaotic dynamics of hybrid light-matter polaritons.

cond-mat.mes-hall↗

arXiv2026

Sub-JEPA: Subspace Gaussian Regularization for Stable End-to-End World Models

Joint-Embedding Predictive Architectures (JEPAs) provide a simpleframework for learning world models by predicting future latent representations.However, JEPA training is subject to a bias-variance tradeoff.Without sufficient structural constraints, excessive representationalvariance causes the model to collapse to trivial solutions.The recent LeWorldModel (LeWM) shows that this issue can be alleviated bysimply constraining latent embeddings with an isotropic Gaussian prior.However, latent representations inherently lie on low-dimensional manifoldswithin a high-dimensional ambient space, and enforcing an isotropic Gaussianprior directly in this ambient space introduces an overly strong bias.In this work, we propose ame, which seeks a favorable operatingpoint on the bias-variance frontier by applying Gaussian constraints inmultiple random subspaces rather than in the originalembedding space.This design relaxes the global constraint while preserving itsanti-collapse effect, leading to a better balance between trainingstability and representation flexibility.Extensive experiments across fourcontinuous-control environments demonstrate that consistentlyoutperforms LeWM with very clear margins.Our method is simple yet effective, and serves as a strong baseline for future JEPA-based world model research.fdefinedeeemodeThe code is available at https://github.com/intcomp/Sub-JEPA.

cs.LG↗

arXiv2025

Diff4Splat: Controllable 4D Scene Generation with Latent Dynamic Reconstruction Models

We introduce Diff4Splat, a feed-forward method that synthesizes controllable and explicit 4D scenes from a single image. Our approach unifies the generative priors of video diffusion models with geometry and motion constraints learned from large-scale 4D datasets. Given a single input image, a camera trajectory, and an optional text prompt, Diff4Splat directly predicts a deformable 3D Gaussian field that encodes appearance, geometry, and motion, all in a single forward pass, without test-time optimization or post-hoc refinement. At the core of our framework lies a video latent transformer, which augments video diffusion models to jointly capture spatio-temporal dependencies and predict time-varying 3D Gaussian primitives. Training is guided by objectives on appearance fidelity, geometric accuracy, and motion consistency, enabling Diff4Splat to synthesize high-quality 4D scenes in 30 seconds. We demonstrate the effectiveness of Diff4Splat across video generation, novel view synthesis, and geometry extraction, where it matches or surpasses optimization-based methods for dynamic scene synthesis while being significantly more efficient.

cs.CV↗

arXiv2025

ID-Crafter: VLM-Grounded Online RL for Compositional Multi-Subject Video Generation

Significant progress has been achieved in high-fidelity video synthesis, yet current paradigms often fall short in effectively integrating identity information from multiple subjects. This leads to semantic conflicts and suboptimal performance in preserving identities and interactions, limiting controllability and applicability. To tackle this issue, we introduce ID-Crafter, a framework for multi-subject video generation that achieves superior identity preservation and semantic coherence. ID-Crafter integrates three key components: (i) a hierarchical identity-preserving attention mechanism that progressively aggregates features at intra-subject, inter-subject, and cross-modal levels; (ii) a semantic understanding module powered by a pretrained Vision-Language Model (VLM) to provide fine-grained guidance and capture complex inter-subject relationships; and (iii) an online reinforcement learning phase to further refine the model for critical concepts. Furthermore, we construct a new dataset to facilitate robust training and evaluation. Extensive experiments demonstrate that ID-Crafter establishes new state-of-the-art performance on multi-subject video generation benchmarks, excelling in identity preservation, temporal consistency, and overall video quality. Project page: https://angericky.github.io/ID-Crafter

cs.CV↗

arXiv2025

Attention Debiasing for Token Pruning in Vision Language Models

Vision-language models (VLMs) typically encode substantially more visual tokens than text tokens, resulting in significant token redundancy. Pruning uninformative visual tokens is therefore crucial for improving computational efficiency, and language-to-vision attention has become a widely used importance criterion for this purpose. However, we find that attention in VLMs is systematically biased. It disproportionately favors tokens appearing later in the sequence, manifesting as over-attention to lower image regions, and assigns inflated scores to semantically empty padding tokens. These behaviors stem from intrinsic recency bias and attention sink effects inherited from large language models (LLMs), and they distort attention-based pruning by preserving irrelevant visual content. To derive a pruning criterion better aligned with semantic relevance, we introduce two lightweight yet effective debiasing techniques that restore the reliability of attention. The first compensates for positional distortions by removing recency-induced attention trends, producing a content-aware and position-agnostic importance measure. The second suppresses attention sink effects by eliminating spurious attention on padding tokens. Our method is model-agnostic, pruning-method-agnostic, and task-agnostic, enabling plug-and-play integration with existing VLM pruning models. Despite its simplicity, our approach consistently delivers strong performance gains. We evaluate our method on ten vision-language benchmarks spanning both image-based and video-based tasks, in comparison with seven state-of-the-art visual token pruning methods and across two representative VLM architectures. Our method achieves substantial performance gains, demonstrating strong effectiveness and generalizability. Our code is available at https://github.com/intcomp/attention-bias.

cs.CV↗

arXiv2025

MoVieS: Motion-Aware 4D Dynamic View Synthesis in One Second

We present MoVieS, a Motion-aware View Synthesis model that reconstructs 4D dynamic scenes from monocular videos in one second. It represents dynamic 3D scenes with pixel-aligned Gaussian primitives and explicitly supervises their time-varying motions. This allows, for the first time, the unified modeling of appearance, geometry and motion from monocular videos, and enables reconstruction, view synthesis and 3D point tracking within a single learning-based framework. By bridging view synthesis with geometry reconstruction, MoVieS enables large-scale training on diverse datasets with minimal dependence on task-specific supervision. As a result, it also naturally supports a wide range of zero-shot applications, such as scene flow estimation and moving object segmentation. Extensive experiments validate the effectiveness and efficiency of MoVieS across multiple tasks, achieving competitive performance while offering several orders of magnitude speedups.

cs.CV↗

arXiv2025

Space-time duality in polariton dynamics

The spatial and temporal dynamics of wave propagation are intertwined. A common manifestation of this duality emerges in the spatial and temporal decay of waves as they propagate through a lossy medium. A complete description of the non-Hermitian wave dynamics in such a lossy system, capturing temporal and spatial decays, necessitates the use of complex-valued frequency and/or wavenumber Eigen-values. Here, we demonstrate that the propagation of polaritons - hybrid light-matter quasiparticles - can be broadly controlled in space and time by temporally shaping their photonic excitation. Using time-domain terahertz near-field nanoscopy, we study plasmon polaritons in bilayer graphene at sub-picosecond time scales. Suppressed spatial decay of polaritons is implemented by temporally engineering the excitation waveform. Polaritonic space-time metrology data agree with our dynamic model. Through the experimental realization and visualization of polaritonic space-time duality, we uncover the effects of the spatio-temporal engineering of wave dynamics; these are applicable to acoustic, photonic, plasmonic, and electronic systems.

physics.optics↗

arXiv2025

PartCrafter: Structured 3D Mesh Generation via Compositional Latent Diffusion Transformers

We introduce PartCrafter, the first structured 3D generative model that jointly synthesizes multiple semantically meaningful and geometrically distinct 3D meshes from a single RGB image. Unlike existing methods that either produce monolithic 3D shapes or follow two-stage pipelines, i.e., first segmenting an image and then reconstructing each segment, PartCrafter adopts a unified, compositional generation architecture that does not rely on pre-segmented inputs. Conditioned on a single image, it simultaneously denoises multiple 3D parts, enabling end-to-end part-aware generation of both individual objects and complex multi-object scenes. PartCrafter builds upon a pretrained 3D mesh diffusion transformer (DiT) trained on whole objects, inheriting the pretrained weights, encoder, and decoder, and introduces two key innovations: (1) A compositional latent space, where each 3D part is represented by a set of disentangled latent tokens; (2) A hierarchical attention mechanism that enables structured information flow both within individual parts and across all parts, ensuring global coherence while preserving part-level detail during generation. To support part-level supervision, we curate a new dataset by mining part-level annotations from large-scale 3D object datasets. Experiments show that PartCrafter outperforms existing approaches in generating decomposable 3D meshes, including parts that are not directly visible in input images, demonstrating the strength of part-aware generative priors for 3D understanding and synthesis. Code and training data will be released.

cs.CV↗

arXiv2025

Information-Guided Identification of Training Data Imprint in (Proprietary) Large Language Models

High-quality training data has proven crucial for developing performant large language models (LLMs). However, commercial LLM providers disclose few, if any, details about the data used for training. This lack of transparency creates multiple challenges: it limits external oversight and inspection of LLMs for issues such as copyright infringement, it undermines the agency of data authors, and it hinders scientific research on critical issues such as data contamination and data selection. How can we recover what training data is known to LLMs? In this work, we demonstrate a new method to identify training data known to proprietary LLMs like GPT-4 without requiring any access to model weights or token probabilities, by using information-guided probes. Our work builds on a key observation: text passages with high surprisal are good search material for memorization probes. By evaluating a model's ability to successfully reconstruct high-surprisal tokens in text, we can identify a surprising number of texts memorized by LLMs.

cs.CL↗

arXiv2025

OmniPhysGS: 3D Constitutive Gaussians for General Physics-Based Dynamics Generation

Recently, significant advancements have been made in the reconstruction and generation of 3D assets, including static cases and those with physical interactions. To recover the physical properties of 3D assets, existing methods typically assume that all materials belong to a specific predefined category (e.g., elasticity). However, such assumptions ignore the complex composition of multiple heterogeneous objects in real scenarios and tend to render less physically plausible animation given a wider range of objects. We propose OmniPhysGS for synthesizing a physics-based 3D dynamic scene composed of more general objects. A key design of OmniPhysGS is treating each 3D asset as a collection of constitutive 3D Gaussians. For each Gaussian, its physical material is represented by an ensemble of 12 physical domain-expert sub-models (rubber, metal, honey, water, etc.), which greatly enhances the flexibility of the proposed model. In the implementation, we define a scene by user-specified prompts and supervise the estimation of material weighting factors via a pretrained video diffusion model. Comprehensive experiments demonstrate that OmniPhysGS achieves more general and realistic physical dynamics across a broader spectrum of materials, including elastic, viscoelastic, plastic, and fluid substances, as well as interactions between different materials. Our method surpasses existing methods by approximately 3% to 16% in metrics of visual quality and text alignment.

cs.CV↗

arXiv2025

Network Dynamics-Based Framework for Understanding Deep Neural Networks

Advancements in artificial intelligence call for a deeper understanding of the fundamental mechanisms underlying deep learning. In this work, we propose a theoretical framework to analyze learning dynamics through the lens of dynamical systems theory. We redefine the notions of linearity and nonlinearity in neural networks by introducing two fundamental transformation units at the neuron level: order-preserving transformations and non-order-preserving transformations. Different transformation modes lead to distinct collective behaviors in weight vector organization, different modes of information extraction, and the emergence of qualitatively different learning phases. Transitions between these phases may occur during training, accounting for key phenomena such as grokking. To further characterize generalization and structural stability, we introduce the concept of attraction basins in both sample and weight spaces. The distribution of neurons with different transformation modes across layers, along with the structural characteristics of the two types of attraction basins, forms a set of core metrics for analyzing the performance of learning models. Hyperparameters such as depth, width, learning rate, and batch size act as control variables for fine-tuning these metrics. Our framework not only sheds light on the intrinsic advantages of deep learning, but also provides a novel perspective for optimizing network architectures and training strategies.

cs.LG↗

arXiv2024

High SNR 3D Imaging from Millimeter-scale Thick Tissues to Cellular Dynamics via Structured Illumination Microscopy

Three-dimensional (3D) fluorescence imaging provides a vital approach for study of biological tissues with intricate structures, and optical sectioning structured illumination microscopy (OS-SIM) stands out for its high imaging speed, low phototoxicity and high spatial resolution. However, OS-SIM faces the problem of low signal-to-noise ratio (SNR) when using traditional decoding algorithms, especially in thick tissues. Here we propose a Hilbert-transform decoding and space domain based high-low (HT-SHiLo) algorithm for noise suppression in OS-SIM. We demonstrate HT-SHiLo algorithm can significantly improve the SNR of optical sectioning images at rapid processing speed, and double the imaging depth in thick tissues. With our OS-SIM system, we achieve high quality 3D images of various biological samples including mouse brains, Drosophila clock neurons, organoids, and live cells. We anticipate that this approach will render OS-SIM a powerful technique for research of cellular organelles or thick tissues in 3D morphology.

physics.optics↗

arXiv2024

CATP: Context-Aware Trajectory Prediction with Competition Symbiosis

Contextual information is vital for accurate trajectory prediction. For instance, the intricate flying behavior of migratory birds hinges on their analysis of environmental cues such as wind direction and air pressure. However, the diverse and dynamic nature of contextual information renders it an arduous task for AI models to comprehend its impact on trajectories and consequently predict them accurately. To address this issue, we propose a ``manager-worker'' framework to unleash the full potential of contextual information and construct CATP model, an implementation of the framework for Context-Aware Trajectory Prediction. The framework comprises a manager model, several worker models, and a tailored training mechanism inspired by competition symbiosis in nature. Taking CATP as an example, each worker needs to compete against others for training data and develop an advantage in predicting specific moving patterns. The manager learns the workers' performance in different contexts and selects the best one in the given context to predict trajectories, enabling CATP as a whole to operate in a symbiotic manner. We conducted two comparative experiments and an ablation study to quantitatively evaluate the proposed framework and CATP model. The results showed that CATP could outperform SOTA models, and the framework could be generalized to different context-aware tasks.

cs.LG↗

arXiv2024

InstructLayout: Instruction-Driven 2D and 3D Layout Synthesis with Semantic Graph Prior

Comprehending natural language instructions is a charming property for both 2D and 3D layout synthesis systems. Existing methods implicitly model object joint distributions and express object relations, hindering generation's controllability. We introduce InstructLayout, a novel generative framework that integrates a semantic graph prior and a layout decoder to improve controllability and fidelity for 2D and 3D layout synthesis. The proposed semantic graph prior learns layout appearances and object distributions simultaneously, demonstrating versatility across various downstream tasks in a zero-shot manner. To facilitate the benchmarking for text-driven 2D and 3D scene synthesis, we respectively curate two high-quality datasets of layout-instruction pairs from public Internet resources with large language and multimodal models. Extensive experimental results reveal that the proposed method outperforms existing state-of-the-art approaches by a large margin in both 2D and 3D layout synthesis tasks. Thorough ablation studies confirm the efficacy of crucial design components.

cs.CV↗