Peng-Jie Li | Searcharxiv

arXiv2026

Understanding Energy Dependent Hadronic Calorimeter Response from a Machine Learning Perspective

To meet the precision requirements of future high-energy physics experiments, improving the energy resolution of hadronic calorimeters remains a critical challenge. This work presents a systematic investigation of hadronic energy reconstruction using machine learning, highlighting the roles of various signal channels, including scintillation light, Cherenkov light, charged particles, and the full three-dimensional topology of hadronic showers in the energy range up to 10 GeV. Throughout this study, detector effects are not taken into account. Under these conditions, the intrinsic resolution of hadronic showers reaches approximately $(10.8\pm0.3)\% / \sqrt{E/GeV}$ when all signal channels and the full 3D shower information are fully utilized. Compared with the traditional signal-summing approach, machine-learning-based reconstruction can significantly improve energy resolution, even under a limited sampling fraction of 10\%, enhancing it from $(57.6\pm3.7)\%/\sqrt{E/GeV}$ to $(34.1\pm2.8)\%/\sqrt{E/GeV}$. These results highlight the critical importance of both multi-channel information and detailed spatial shower features in hadronic energy reconstruction, and demonstrate the substantial potential of combining high-granularity and dual-readout calorimeter designs with machine-learning-based reconstruction techniques for future experiments.

hep-ex↗

arXiv2025

Attention-Driven Multimodal Alignment for Long-term Action Quality Assessment

Long-term action quality assessment (AQA) focuses on evaluating the quality of human activities in videos lasting up to several minutes. This task plays an important role in the automated evaluation of artistic sports such as rhythmic gymnastics and figure skating, where both accurate motion execution and temporal synchronization with background music are essential for performance assessment. However, existing methods predominantly fall into two categories: unimodal approaches that rely solely on visual features, which are inadequate for modeling multimodal cues like music; and multimodal approaches that typically employ simple feature-level contrastive fusion, overlooking deep cross-modal collaboration and temporal dynamics. As a result, they struggle to capture complex interactions between modalities and fail to accurately track critical performance changes throughout extended sequences. To address these challenges, we propose the Long-term Multimodal Attention Consistency Network (LMAC-Net). LMAC-Net introduces a multimodal attention consistency mechanism to explicitly align multimodal features, enabling stable integration of visual and audio information and enhancing feature representations. Specifically, we introduce a multimodal local query encoder module to capture temporal semantics and cross-modal relations, and use a two-level score evaluation for interpretable results. In addition, attention-based and regression-based losses are applied to jointly optimize multimodal alignment and score fusion. Experiments conducted on the RG and Fis-V datasets demonstrate that LMAC-Net significantly outperforms existing methods, validating the effectiveness of our proposed approach.

cs.CV↗