4arXiv cs.AI (Artificial Intelligence)·
RMISC: Large-scale real-world multivariate corpus for time series foundation model pretraining
Researchers introduce RMISC, an openly accessible corpus of ~200 datasets and 142 billion time points of real-world multivariate time series data, designed to address the gap between synthetic and real-world pretraining data for time series foundation models (TSFMs). Four advanced TSFMs are pretrained on univariate, synthetic multivariate, and real-world multivariate data and evaluated on zero-shot generalization benchmarks. Results show that real-world multivariate pretraining data consistently improves generalization for both univariate and multivariate TSFMs. The work provides both a reusable dataset resource and empirical evidence on the synthetic-vs-real data question for time series modeling.
Related guides (2)
Related events (8)
TiRex-2: xLSTM-based time series foundation model for multivariate and streaming forecasting
TiRex-2 is a recurrent xLSTM-based time series foundation model that extends the univariate TiRex to multivariate forecasting with past and future covariates, while supporting streaming inference at constant per-patch cost. The model uses a bidirectional time mixer and asymmetric grouped-attention variate mixer to handle future-known covariates without violating causality over target variables. A synthetic coupling pipeline enables scalable multivariate pretraining from univariate corpora. TiRex-2 claims state-of-the-art zero-shot performance on GIFT-Eval and fev-bench benchmarks with 38.4M–82.5M parameters depending on mode.
FusionRS: Large-scale RGB-infrared-text dataset for dual-modal remote sensing vision-language models
Researchers introduce FusionRS, the first large-scale dataset pairing RGB and infrared remote sensing images with both conventional and IR-aware text captions, designed to support dual-modal vision-language learning. The dataset is constructed by translating public RGB remote sensing images into infrared-style counterparts using image translation. Using FusionRS, the authors train CLIP-style alignment models and fine-tune generative VLMs, demonstrating improvements in RGB-IR alignment, infrared-to-text retrieval, and dual-modal captioning over RGB-only baselines. The work addresses a gap in multimodal remote sensing foundation models by providing modality-specific textual supervision for infrared imagery.
TimesFM: Google Research Pretrained Time-Series Foundation Model
TimesFM is a pretrained foundation model developed by Google Research specifically for time-series forecasting tasks. The repository has accumulated over 20,000 GitHub stars with 99 new stars today, indicating sustained community interest. It represents Google Research's effort to apply the foundation model paradigm to time-series data rather than language or vision.
VisAnomReasoner: Efficient VLM for Time-Series Anomaly Detection via VisAnomBench
Researchers introduce VisAnomBench, a curated benchmark augmenting public time-series anomaly datasets with natural-language rationales generated and selected from multiple large VLMs using task-specific rewards. Fine-tuning on this benchmark produces VisAnomReasoner, a parameter-efficient vision-language model that outperforms all baselines by at least 21.23 and 23.87 percentage points in precision and F1 on VisAnomBench. Cross-benchmark evaluation on TSB-AD-U shows further generalization gains of 9.57 and 13.39 percentage points in precision and F1, respectively.
OpenMedReason: Large-scale multimodal medical reasoning corpus with 450K instances for clinical VLM training
Researchers introduce OpenMedReason, a 450K-instance open multimodal medical reasoning corpus with reasoning traces derived from human-authored biomedical literature rather than synthetic chains of thought. The dataset covers diverse medical imaging modalities and is paired with OpenMedReason-Bench, a held-out benchmark evaluating LVLMs on perception, medical knowledge, and rationale axes. Training with OpenMedReason yields a 20% average VQA accuracy improvement over base models and achieves performance within 4.2% of leading comparable-scale medical VLMs. Both the dataset and code are publicly released.
Stanford EDGAR Filings Dataset: 152B-token open corpus of SEC filings for LLM pretraining
Stanford researchers introduce the Stanford EDGAR Filings Dataset (SEFD), an open reconstruction of SEC filings into layout-faithful MultiMarkdown, releasing a 152B-token initial snapshot with a larger 550B-token archive described. The dataset targets the growing scarcity of high-quality long-context pretraining data, with less than 0.1% overlap with Common Crawl-derived corpora. Two derived benchmarks are also introduced: EDGAR-Forecast for filing-grounded numerical forecasting and EDGAR-OCR for complex financial table transcription. The work addresses a real gap in open long-context training data outside narrow domains like code.
SARLO-80: Large-scale VHR SAR-optical-text dataset for multimodal foundation model training
Researchers from ONERA release SARLO-80, a dataset of 119,566 triplets combining very-high-resolution complex SAR imagery, aligned optical patches, and natural-language captions covering 257 locations across 72 countries. The dataset is built from Umbra spotlight acquisitions standardized to an 80cm slant-range grid, with three caption variants per sample to support vision-language training and evaluation. It addresses a recognized gap in SAR-optical multimodal resources, which have historically been limited to low-resolution intensity-only products. The dataset and preprocessing code are publicly released on Hugging Face Hub.
Foundation Model for Wearable Health Data Pretrained on 1 Trillion Minutes from 5 Million Participants
Researchers propose a large-scale foundation model for wearable health data, pretrained on over one trillion minutes of unlabeled sensor signals from five million participants. The model demonstrates systematic performance improvements across 35 health prediction tasks spanning cardiovascular, metabolic, sleep, and mental health domains, with joint scaling of model capacity and data volume. A 'classroom' of LLM agents autonomously searches downstream predictive head configurations, and the resulting embeddings are integrated into a Personal Health Agent validated by 1,860 clinician ratings. The work establishes label-efficient few-shot learning and generative capabilities for daily health metric estimation.