WilLaGS: Latent-Conditional 3D Appearance Fields for Robust Gaussian Splatting In-the-Wild
为解决野外场景中3D高斯点绘的多视角一致性问题，提出WilLaGS框架，通过生成式外观模型和自监督感知掩模机制实现鲁棒的3D场景重建与渲染。
0 citationsRead paper
Selected work
Representative Papers
Registration-Free Hyperspectral Reconstruction from RGB via a Permutation-Invariant Gram-Matrix Principle
This study addresses the challenge of hyperspectral image fusion under unregistered conditions and unknown camera response functions by proposing a supervised framework based on permutation-invariant Gram matrices. Leveraging the permutation invariance of abundance map Gram matrices, the method integrates a residual spectral super-resolution network with a random pixel permutation validation strategy to enable direct RGB-to-high-resolution hyperspectral reconstruction without spatial registration, camera response functions, or paired data. Experiments demonstrate that this approach achieves accuracy comparable to conventional assumption-dependent methods across multiple scene benchmarks while maintaining robustness when such assumptions fail. Notably, its performance advantage stems from fundamental principled innovation rather than loss function tuning, offering a reliable solution for real-world scenarios where standard calibration and alignment prerequisites are unavailable.
0 citationsRead paper
Journal Impact Factor and Federal Reserve Monetary Policy: An Econometric Analysis Based on 1975-2026
This study investigates the long-term impact of U.S. Federal Reserve monetary policy on academic journal impact factors and its underlying macroeconomic transmission mechanisms. Using time-series data from 1975 to 2026, the authors estimate three nested ordinary least squares (OLS) models—a linear baseline specification, a version controlling for time trends, and a log-transformed variant—and report, for the first time, a significant negative relationship between real interest rates and impact factors during periods of quantitative easing. Specifically, between 2001 and 2020, a one-percentage-point decline in real interest rates was associated with an average 6.9% increase in impact factors. The full-sample model achieves an adjusted R² of 0.893. The paper introduces a novel conceptual framework termed “financialization of academic capital,” elucidating how monetary policy indirectly reshapes scholarly evaluation systems through capital allocation channels.
0 citationsRead paper
DA-Occ: Efficient 3D Voxel Occupancy Prediction via Directional 2D for Geometric Structure Preservation
To address the dual requirements of accuracy and inference speed for 3D voxel occupancy prediction in autonomous driving, this paper proposes a lightweight and efficient directional 2D feature modeling framework. The method preserves vertical geometric structure via directional feature slicing, recovers height cues by fusing multi-view 2D features using a novel directional attention mechanism, and incorporates geometric-aware design in the bird’s-eye view (BEV) space—eliminating the computational overhead of explicit 3D convolutions. Evaluated on Occ3D-nuScenes, our approach achieves 39.3% mIoU at 27.7 FPS on GPU and 14.8 FPS on edge devices, substantially outperforming existing real-time methods. It establishes a new Pareto-optimal trade-off between accuracy and efficiency while maintaining 3D structural integrity.
0 citationsRead paper
ICH-Qwen: A Large Language Model Towards Chinese Intangible Cultural Heritage
China’s intangible cultural heritage (ICH) faces severe challenges—including transmission discontinuity and skill attrition—amid rapid modernization. Existing large language models (LLMs) lack domain-specific adaptation for ICH, limiting their applicability in digital humanities and heritage preservation. To address this, we introduce the first Chinese LLM dedicated to Chinese ICH: built upon the Qwen architecture, it integrates domain-specific pretraining on ICH corpora, synthetic data augmentation tailored to ICH knowledge, supervised fine-tuning, and explicit knowledge alignment. This model achieves the first systematic deep semantic modeling of ICH within LLMs. Empirical evaluation demonstrates substantial improvements over general-purpose baselines across key tasks—including ICH question answering, generative description of traditional craftsmanship, and simulated dialogues with heritage bearers. The work provides a deployable, scalable technical framework and methodological paradigm for intelligent ICH preservation and digital humanities research.
0 citationsRead paper
Recent publications
Latest Papers
WilLaGS: Latent-Conditional 3D Appearance Fields for Robust Gaussian Splatting In-the-Wild
为解决野外场景中3D高斯点绘的多视角一致性问题，提出WilLaGS框架，通过生成式外观模型和自监督感知掩模机制实现鲁棒的3D场景重建与渲染。
0 citationsRead paper
Registration-Free Hyperspectral Reconstruction from RGB via a Permutation-Invariant Gram-Matrix Principle
This study addresses the challenge of hyperspectral image fusion under unregistered conditions and unknown camera response functions by proposing a supervised framework based on permutation-invariant Gram matrices. Leveraging the permutation invariance of abundance map Gram matrices, the method integrates a residual spectral super-resolution network with a random pixel permutation validation strategy to enable direct RGB-to-high-resolution hyperspectral reconstruction without spatial registration, camera response functions, or paired data. Experiments demonstrate that this approach achieves accuracy comparable to conventional assumption-dependent methods across multiple scene benchmarks while maintaining robustness when such assumptions fail. Notably, its performance advantage stems from fundamental principled innovation rather than loss function tuning, offering a reliable solution for real-world scenarios where standard calibration and alignment prerequisites are unavailable.
0 citationsRead paper
Journal Impact Factor and Federal Reserve Monetary Policy: An Econometric Analysis Based on 1975-2026
This study investigates the long-term impact of U.S. Federal Reserve monetary policy on academic journal impact factors and its underlying macroeconomic transmission mechanisms. Using time-series data from 1975 to 2026, the authors estimate three nested ordinary least squares (OLS) models—a linear baseline specification, a version controlling for time trends, and a log-transformed variant—and report, for the first time, a significant negative relationship between real interest rates and impact factors during periods of quantitative easing. Specifically, between 2001 and 2020, a one-percentage-point decline in real interest rates was associated with an average 6.9% increase in impact factors. The full-sample model achieves an adjusted R² of 0.893. The paper introduces a novel conceptual framework termed “financialization of academic capital,” elucidating how monetary policy indirectly reshapes scholarly evaluation systems through capital allocation channels.
0 citationsRead paper
DA-Occ: Efficient 3D Voxel Occupancy Prediction via Directional 2D for Geometric Structure Preservation
To address the dual requirements of accuracy and inference speed for 3D voxel occupancy prediction in autonomous driving, this paper proposes a lightweight and efficient directional 2D feature modeling framework. The method preserves vertical geometric structure via directional feature slicing, recovers height cues by fusing multi-view 2D features using a novel directional attention mechanism, and incorporates geometric-aware design in the bird’s-eye view (BEV) space—eliminating the computational overhead of explicit 3D convolutions. Evaluated on Occ3D-nuScenes, our approach achieves 39.3% mIoU at 27.7 FPS on GPU and 14.8 FPS on edge devices, substantially outperforming existing real-time methods. It establishes a new Pareto-optimal trade-off between accuracy and efficiency while maintaining 3D structural integrity.
0 citationsRead paper
ICH-Qwen: A Large Language Model Towards Chinese Intangible Cultural Heritage
China’s intangible cultural heritage (ICH) faces severe challenges—including transmission discontinuity and skill attrition—amid rapid modernization. Existing large language models (LLMs) lack domain-specific adaptation for ICH, limiting their applicability in digital humanities and heritage preservation. To address this, we introduce the first Chinese LLM dedicated to Chinese ICH: built upon the Qwen architecture, it integrates domain-specific pretraining on ICH corpora, synthetic data augmentation tailored to ICH knowledge, supervised fine-tuning, and explicit knowledge alignment. This model achieves the first systematic deep semantic modeling of ICH within LLMs. Empirical evaluation demonstrates substantial improvements over general-purpose baselines across key tasks—including ICH question answering, generative description of traditional craftsmanship, and simulated dialogues with heritage bearers. The work provides a deployable, scalable technical framework and methodological paradigm for intelligent ICH preservation and digital humanities research.
0 citationsRead paper