June 10, 2025 – Kavli Institute Pre-Print Publications

TTrace: Lightweight Error Checking and Diagnosis for Distributed Training

Posted by klaurent June 10, 2025Peking

Kavli Affiliate: Li Xin Li| First 5 Authors: [#item_custom_name[1, [#item_custom_name[2, [#item_custom_name[3, [#item_custom_name[4, [#item_custom_name[5| Summary:Distributed training is essential for scaling the training of large neural network models, such as large language models (LLMs), across thousands of GPUs. However, the complexity of distributed training programs makes them particularly prone to silent bugs, which do not produce explicit […]

Continue..

Mitigating antenna gain errors with HyFoReS in CHIME simulations

Posted by dbos June 10, 2025June 23, 2025MIT

Kavli Affiliate: Kiyoshi W. Masui | First 5 Authors: Haochen Wang, Panupong Phoompuang, Kiyoshi W. Masui, Arnab Chakraborty, Simon Foreman | Summary: Hybrid Foreground Residual Subtraction (HyFoReS) is a new family of algorithms designed to remove systematics-induced foreground contamination for 21-cm intensity mapping data. Previously, the algorithm was shown to be effective in mitigating beam […]

Continue..

MagCache: Fast Video Generation with Magnitude-Aware Cache

Posted by klaurent June 10, 2025Berkeley

Kavli Affiliate: Feng Wang | First 5 Authors: Zehong Ma, Zehong Ma, , , | Summary: Existing acceleration techniques for video diffusion models often rely on uniform heuristics or time-embedding variants to skip timesteps and reuse cached features. These approaches typically require extensive calibration with curated prompts and risk inconsistent outputs due to prompt-specific overfitting. […]

Continue..

MagCache: Fast Video Generation with Magnitude-Aware Cache

Posted by dbos June 10, 2025June 23, 2025Berkeley

Kavli Affiliate: Feng Wang | First 5 Authors: Zehong Ma, Longhui Wei, Feng Wang, Shiliang Zhang, Qi Tian | Summary: Existing acceleration techniques for video diffusion models often rely on uniform heuristics or time-embedding variants to skip timesteps and reuse cached features. These approaches typically require extensive calibration with curated prompts and risk inconsistent outputs […]

Continue..

Identifying vortex lattice in type-II superconductors via the dynamic magnetostrictive effect

Posted by dbos June 10, 2025June 23, 2025UCAS

Kavli Affiliate: Long Zhang | First 5 Authors: Peipei Lu, Mengju Yuan, Jing Zhang, Qiang Gao, Shuang Liu | Summary: In type-I superconductors, zero electrical resistivity and perfect diamagnetism define two fundamental criteria for superconducting behavior. In contrast, type-II superconductors exhibit more complex mixed-state physics, where magnetic flux penetrates the material above the lower critical […]

Continue..

High-precision Beam Optics Calculation of the HIAF-BRing Using Measured Fields

Posted by klaurent June 10, 2025Peking

Kavli Affiliate: Ke Wang | First 5 Authors: , , , , | Summary: The construction of the High Intensity heavy ion Accelerator Facility (HIAF) has been completed, with current efforts focused on subsystem commissioning. Beam commissioning is scheduled for autumn 2025, marking a critical milestone in the HIAF project. This paper presents high-precision optics […]

Continue..

High-precision Beam Optics Calculation of the HIAF-BRing Using Measured Fields

Posted by dbos June 10, 2025Peking

Kavli Affiliate: Ke Wang | First 5 Authors: Ke Wang, Li-Na Sheng, Geng Wang, Wei-Ping Chai, You-Jin Yuan | Summary: The construction of the High Intensity heavy ion Accelerator Facility (HIAF) has been completed, with current efforts focused on subsystem commissioning. Beam commissioning is scheduled for autumn 2025, marking a critical milestone in the HIAF […]

Continue..

Pureformer-VC: Non-parallel Voice Conversion with Pure Stylized Transformer Blocks and Triplet Discriminative Training

Posted by dbos June 10, 2025June 23, 2025Tokyo

Kavli Affiliate: Jia Liu | First 5 Authors: Wenhan Yao, Fen Xiao, Xiarun Chen, Jia Liu, YongQiang He | Summary: As a foundational technology for intelligent human-computer interaction, voice conversion (VC) seeks to transform speech from any source timbre into any target timbre. Traditional voice conversion methods based on Generative Adversarial Networks (GANs) encounter significant […]

Continue..