Controlled Study Identifies Key Design Factors for Stable Autoregressive Forecasting of Seismic Wavefields
Researchers conducted a controlled ablation study on SeismoGPT, an autoregressive model for forecasting seismograms, to understand what prevents error accumulation over long prediction horizons. Multi-token prediction emerged as the dominant stabilizing factor, with context length and loss function design playing secondary but important roles. The findings have implications for improving autoregressive models for oscillatory physical signals like seismic data and gravitational waves.
A new study published on arXiv examines why autoregressive sequence models struggle with long-horizon forecasting of oscillatory physical signals such as seismograms. The core problem is error accumulation: as models generate predictions step-by-step and feed their own outputs back as inputs, small errors compound into phase drift that standard pointwise metrics fail to capture. Using synthetic three-component seismograms as a controlled testbed, researchers systematically ablated design choices in the SeismoGPT model through free-running rollout experiments with paired significance tests. Multi-token prediction (predicting multiple timesteps simultaneously) accounted for nearly all performance gains over single-token baselines, while a horizon-embedding hybrid prediction head and cross-horizon STFT-magnitude coherence loss provided smaller but consistent improvements. The study identified a sharp performance threshold at a context ratio near one (approximately the P-S interval of observed seismic signals), below which generalization collapses. A remaining challenge is polarity inversion, which magnitude-based spectral losses cannot penalize, suggesting phase-aware objectives as a promising direction.
What's missing
The study does not discuss computational costs or inference time comparisons between different design choices, nor does it address how findings generalize to real (non-synthetic) seismograms or other types of oscillatory wavefields beyond the controlled synthetic testbed.
What different sources said
- arXiv cs.LGCenter
When Do Autoregressive Sequence Models Forecast Physical Wavefields? A Controlled Study on Synthetic Seismograms
Related
Gut Bacteria Enzyme Found to Break Down Heat-Processed Food Compounds, Producing Novel Biogenic Amines
Researchers have discovered that an enzyme in common gut bacteria can degrade N-epsilon-carboxymethyllysine (CML), a compound formed during thermal food processing, producing previously unknown biogenic amines. The enzyme, ornithine decarboxylase SpeC from enterobacteria, acts on CML and related modified lysine derivatives through a low-level 'underground' catalytic activity. This finding suggests a previously unrecognized communication axis between thermally processed dietary compounds and gut microbial physiology, with potential implications for host health.
Full-Length Gene Sequencing Reveals Two Distinct Bacterial Communities in Black-Legged Ticks Expanding Into Canada
Researchers used Oxford Nanopore full-length 16S rRNA gene sequencing to characterize the microbiome of Ixodes scapularis black-legged ticks collected in Nova Scotia, Canada, distinguishing between tick-adapted bacteria and environmentally acquired bacteria. The study comes as I. scapularis — the primary vector of Lyme disease — is rapidly expanding northward into Canada due to climate change. The findings suggest that environmentally derived bacteria in tick microbiomes are not mere contamination, which has implications for how tick microbiome data is collected and interpreted across surveillance studies.
Study Identifies Metabolic Link Between Cell Envelope Stress and Biofilm Formation in Bacteria
Researchers have discovered that the metabolite acetyl-CoA directly inhibits enzymes that degrade the bacterial signaling molecule c-di-GMP, connecting cell envelope biosynthesis stress to biofilm formation in Pseudomonas aeruginosa. The study found that sub-inhibitory concentrations of antibiotics targeting early peptidoglycan biosynthesis — but not other antibiotic classes — elevate c-di-GMP levels by reducing phosphodiesterase activity, with acetyl-CoA competing for the enzyme active site. Because the relevant enzyme domain is broadly conserved across bacterial species, this checkpoint mechanism may be widespread and could have implications for understanding antibiotic-induced biofilm responses.