Selected work

Representative Papers

GENEB: Why Genomic Models Are Hard to Compare

Jun 03, 2026

This study addresses the lack of standardized evaluation protocols that hinder fair assessment of genomic foundation models’ performance and generalization. To this end, the authors introduce GENEB, a large-scale diagnostic benchmark that systematically evaluates frozen representations from 40 models across 100 tasks under a unified probing protocol, spanning 13 functional categories and supporting few-shot settings. This framework enables, for the first time, category-aware, fine-grained, and controllable multidimensional comparisons, revealing the instability of aggregate leaderboards and inherent trade-offs across tasks. Key findings indicate substantial variation in model rankings across functional categories, limited and inconsistent gains from increased model scale, and a more decisive influence of architectural design and alignment between pretraining data and downstream tasks than parameter count alone.

0 citationsRead paper

Few-Shot Synthetic Data Generation with Diffusion Models for Downstream Vision Tasks

May 12, 2026

This work addresses the challenge of class imbalance in domains such as medical imaging and industrial defect inspection, where positive (rare-class) samples are scarce. The authors propose a lightweight data augmentation approach leveraging pre-trained diffusion models, requiring only 20–50 real rare-class images to generate high-quality synthetic data through efficient fine-tuning of LoRA adapters. The study systematically investigates the optimal mixing ratio between synthetic and real data for downstream tasks. Notably, this is the first method to integrate LoRA-based fine-tuning with diffusion models for data augmentation, significantly improving recall and F1 scores for rare classes. Extensive experiments on chest X-ray and magnetic tile defect datasets demonstrate the method’s effectiveness and cross-domain scalability.

0 citationsRead paper

Weyl-Heisenberg Transform Capabilities in JPEG Compression Standard

Nov 14, 2025

To address the blocking artifacts and insufficient frequency-domain localization inherent in JPEG’s discrete cosine transform (DCT), this paper proposes a novel image compression method based on the real two-dimensional discrete Weyl–Heisenberg transform (DWHT). For the first time, the Weyl–Heisenberg orthogonal basis—exhibiting joint time-frequency compact localization—is integrated into the JPEG framework as a drop-in replacement for DCT, preserving format compatibility while substantially improving energy compaction and decorrelation. The method incorporates adaptive quantization, optimized entropy coding, and newly designed metrics for compression efficiency evaluation and distortion measurement. Experimental results demonstrate that, at equivalent subjective quality, the proposed approach achieves an average PSNR gain of 1.2–2.8 dB over JPEG, with notably superior detail preservation at low bitrates (<0.5 bpp) and significantly enhanced compression efficiency.

0 citationsRead paper

Recent publications

Latest Papers

GENEB: Why Genomic Models Are Hard to Compare

Jun 03, 2026

This study addresses the lack of standardized evaluation protocols that hinder fair assessment of genomic foundation models’ performance and generalization. To this end, the authors introduce GENEB, a large-scale diagnostic benchmark that systematically evaluates frozen representations from 40 models across 100 tasks under a unified probing protocol, spanning 13 functional categories and supporting few-shot settings. This framework enables, for the first time, category-aware, fine-grained, and controllable multidimensional comparisons, revealing the instability of aggregate leaderboards and inherent trade-offs across tasks. Key findings indicate substantial variation in model rankings across functional categories, limited and inconsistent gains from increased model scale, and a more decisive influence of architectural design and alignment between pretraining data and downstream tasks than parameter count alone.

0 citationsRead paper

Few-Shot Synthetic Data Generation with Diffusion Models for Downstream Vision Tasks

May 12, 2026

This work addresses the challenge of class imbalance in domains such as medical imaging and industrial defect inspection, where positive (rare-class) samples are scarce. The authors propose a lightweight data augmentation approach leveraging pre-trained diffusion models, requiring only 20–50 real rare-class images to generate high-quality synthetic data through efficient fine-tuning of LoRA adapters. The study systematically investigates the optimal mixing ratio between synthetic and real data for downstream tasks. Notably, this is the first method to integrate LoRA-based fine-tuning with diffusion models for data augmentation, significantly improving recall and F1 scores for rare classes. Extensive experiments on chest X-ray and magnetic tile defect datasets demonstrate the method’s effectiveness and cross-domain scalability.

0 citationsRead paper

Weyl-Heisenberg Transform Capabilities in JPEG Compression Standard

Nov 14, 2025

To address the blocking artifacts and insufficient frequency-domain localization inherent in JPEG’s discrete cosine transform (DCT), this paper proposes a novel image compression method based on the real two-dimensional discrete Weyl–Heisenberg transform (DWHT). For the first time, the Weyl–Heisenberg orthogonal basis—exhibiting joint time-frequency compact localization—is integrated into the JPEG framework as a drop-in replacement for DCT, preserving format compatibility while substantially improving energy compaction and decorrelation. The method incorporates adaptive quantization, optimized entropy coding, and newly designed metrics for compression efficiency evaluation and distortion measurement. Experimental results demonstrate that, at equivalent subjective quality, the proposed approach achieves an average PSNR gain of 1.2–2.8 dB over JPEG, with notably superior detail preservation at low bitrates (<0.5 bpp) and significantly enhanced compression efficiency.

0 citationsRead paper