arXiv subjects

Xiaoyun Yu

Publications and source records attributed to Xiaoyun Yu.

5 recordsLinked to original sources

arXiv2026

Decoupling Logical Masks from GPU Execution for Dynamic Block-Sparse Attention

Attention computation makes inference expensive in video diffusion transformers (vDiTs), which generate videos through iterative denoising. Block-sparse attention (BSA) reduces this cost by computing only blocks selected by a logical mask, which specifies attention interactions to compute. However, coupling logical block geometry to execution choices limits adaptation to varying masks and graphics processing units (GPUs), while runtime kernel specialization can incur preparation overhead that outweighs execution time savings. We present Tessera, a specialized runtime for dynamic BSA that decouples logical masks from GPU execution while preserving specified attention interactions. Its physical mapping layer retains, combines, or subdivides logical attention blocks into physical tiles suited to different attention mask shapes and GPU architectures. Its task organization layer groups and schedules tiles within GPU tasks to reuse data, expose parallelism, and overlap data movement with computation. Finally, profile-guided regime selection enables low- overhead execution plan selection through a lookup table constructed from offline profiling. We implement Tessera with specialized CUDA kernels supporting four NVIDIA GPU generations. Evaluated on 2,315 real attention masks and industrial video diffusion models, Tessera achieves up to 6.79x BSA request speedup over baseline systems in the evaluated video diffusion models.

cs.AR↗

arXiv2026

MUSE: Dependency-Aware Adaptation of a Frozen Vision Backbone for Multivariate Time Series Forecasting

Multivariate time-series forecasting is essential to many real-world applications. Recent large vision models (LVMs) offer a promising paradigm by transferring cross-domain visual priors to time-series forecasting. However, existing LVM-based methods face two key challenges: balancing independent visual representation spaces with cross-variable dependency modeling, and adapting vision backbones pretrained on natural images to the distinct temporal semantics of time-series images. To address these challenges, we propose MUSE, a dependency-aware adaptation framework built on a fully frozen pretrained MAE. First, the Variable Context Refinement Module (VCR) aggregates shared temporal information within each variable and models cross-variable contextual dependencies while preserving independent visual spaces. Second, the Temporal-Periodic Refinement Module (TPR) performs lightweight refinement at different encoder depths and explicitly models across-period temporal dependencies and within-period periodic dependencies. The two modules independently produce forecasts, which are fused through a learnable prediction-level gate. Experiments on 10 real-world datasets demonstrate that MUSE achieves state-of-the-art performance.

cs.LG↗

arXiv2026

MEMTS: Internalizing Domain Knowledge via Parameterized Memory for Retrieval-Free Domain Adaptation of Time Series Foundation Models

While Time Series Foundation Models (TSFMs) have demonstrated exceptional performance in generalized forecasting, their performance often degrades significantly when deployed in real-world vertical domains characterized by temporal distribution shifts and domain-specific periodic structures. Current solutions are primarily constrained by two paradigms: Domain-Adaptive Pretraining (DAPT), which improves short-term domain fitting but frequently disrupts previously learned global temporal patterns due to catastrophic forgetting; and Retrieval-Augmented Generation (RAG), which incorporates external knowledge but introduces substantial retrieval overhead. This creates a severe scalability bottleneck that fails to meet the high-efficiency requirements of real-time stream processing. To break this impasse, we propose Memory for Time Series (MEMTS), a lightweight and plug-and-play method for retrieval-free domain adaptation in time series forecasting. The key component of MEMTS is a Knowledge Persistence Module (KPM), which internalizes domain-specific temporal dynamics, such as recurring seasonal patterns and trends into a compact set of learnable latent prototypes. In doing so, it transforms fragmented historical observations into continuous, parameterized knowledge representations. This paradigm shift enables MEMTS to achieve accurate domain adaptation with constant-time inference and near-zero latency, while effectively mitigating catastrophic forgetting of general temporal patterns, all without requiring any architectural modifications to the frozen TSFM backbone. Extensive experiments on multiple datasets demonstrate the SOTA performance of MEMTS.

cs.LG↗

arXiv2018

Critical behaviors of half-metallic ferromagnet Co3Sn2S2

We have investigated the critical behavior of a shandite-type half-metal ferromagnet Co3Sn2S2. It exhibits a second-order paramagnetic-ferromagnetic phase transition with TC = 174 K. To investigate the nature of the magnetic phase transition, a detailed critical exponent study has been performed. The critical components beta, gamma, and delta determined using the modified Arrott plot, the Kouvel-Fisher method as well as the critical isotherm analysis are match reasonably well and follow the scaling equation, confirming that the exponents are unambiguous and intrinsic to the material. The determined exponents of Co3Sn2S2 deviates from theoretical estimated short-range universal models. Instead, Co3Sn2S2 exhibits long-range order in the nature of magnetic interaction with the spin decay as J(r) ~ 1/r^[-(d + sigma)] with sigma = 1.28.

cond-mat.str-el↗

arXiv2016

Enhancing thermal stability of solution-processed small molecule semiconductor thin films using a flexible linker approach

Solution-processed organic photovoltaics (OPV) have recently reached the target 10% power conversion efficiency expected to signal their viable commercialization as an inexpensive and scalable energy conversion technology. However, obtaining devices with suitable long-term stability remains an unsolved challenge. Here we present a new strategy to improve the thermal stability of small-molecule-based bulk-heterojunction OPVs by including a custom additive specifically designed to interact with the device active layer components. Our results indicate that active layer degradation under continuous thermal stress can be inhibited due to the formation of more robust thin film microstructure with the additive present. Since our additive employs the identical semiconductor core used in the active layer, but linked by aliphatic chains into a flexible polymer, this straightforward strategy can reasonably be applied to stabilize a wide variety of semiconducting small molecules in solution-processed molecular OPVs, transistors and light emitting diodes.

cond-mat.mtrl-sci↗