SPADE: New Autoregressive Transformer Method for High-Granularity Calorimeter Simulation
Researchers have introduced SPADE (Split-and-Delay Embeddings), an autoregressive transformer designed to handle sequences where individual tokens contain multiple features by embedding them independently rather than jointly. The method was tested on point-cloud calorimeter shower generation in the ILD detector, where it matched state-of-the-art performance on photon showers and significantly outperformed previous VQ-VAE-based approaches. The technique is broadly applicable to any generative task involving multi-feature tokens and could enable large language model-style pretraining workflows for higher-dimensional scientific data.
SPADE introduces a novel approach to autoregressive transformer design by treating multi-feature tokens differently from conventional methods. Rather than embedding all features of a token together, SPADE embeds each feature independently and staggers them temporally, allowing the standard self-attention mechanism to learn correlations within tokens. The method was evaluated on a challenging physics application: generating calorimeter shower patterns in the highly granular ILD detector. Results demonstrate that SPADE achieves competitive performance with the state-of-the-art AllShowers model for photon shower generation while substantially improving upon its immediate predecessor, OmniJet-αC. The authors emphasize that this mechanism extends beyond particle physics applications to any generative modeling task involving tokens with multiple features, potentially opening new pathways for applying large language model-style pretraining approaches to scientific and engineering domains with higher-dimensional data.
What different sources said
- arXiv cs.LGCenter
SPADE: Split-and-Delay Embeddings for Autoregressive High-Granularity Calorimeter Simulation
Related
Genetic Drift, Not Selection, Drives Rapid Feather Color Evolution in Island Bird Radiation
A new study of an island bird radiation found that rapid evolution of feather coloration is driven primarily by genetic drift in small populations rather than sexual or ecological selection. The research integrated whole-genome data with detailed plumage measurements across complete species sampling to test whether signaling trait evolution correlates with speciation rates. The findings suggest that neutral demographic processes play a central role in generating phenotypic diversity during island radiations, challenging assumptions about the mechanisms driving rapid evolution.
New AI Model Improves Prediction of Therapeutic Peptide Function from Protein Sequences
Researchers developed a lightweight CNN classifier that predicts whether peptide sequences have therapeutic properties, trained on a database of 54,655 peptides across 48 functional categories. The model uses a novel negative sampling strategy to reduce false positive rates from over 60% in previous approaches to 2.1%. This advancement could accelerate drug discovery by enabling faster computational screening of peptide candidates before expensive experimental testing.
Study Shows Different Metabolic Stress Models Produce Distinct Effects on Human Neuronal Networks
Researchers tested three common in vitro metabolic stress models on human-derived neuronal networks and found each produced different patterns of neuronal activity and cell damage. The models tested were hypoxia alone, oxygen-glucose deprivation (OGD), and hypoxia combined with glutamate exposure. The findings suggest that choice of experimental model significantly affects results and that combining electrophysiological and structural analyses is important for accurately assessing metabolic stress in stroke research.