Four new research papers explore advancements in Time-Series Foundation Models (TSFMs), focusing on improving their forecasting capabilities. One paper introduces a method to evaluate TSFMs on sparse event data, finding that current models offer limited improvement over simpler references and suggesting the use of event supervision. Another paper proposes a loss-guided pretraining data selection framework to optimize the learning signal from large datasets. A third paper presents Latent Inference-Time Guidance, an adaptive ensembling approach for TSFMs that combines forecasts through a time-dependent latent space. The fourth paper introduces RACE, a framework for test-time adaptation that improves TSFM performance in neighbor-rich forecasting scenarios by aligning and aggregating evidence from related series. AI
IMPACT These advancements could lead to more accurate and efficient forecasting in various domains, from financial markets to resource management.
RANK_REASON Cluster consists of multiple academic papers on a specific AI research topic.
- alphaXiv
- arXiv
- CatalyzeX
- CORE Recommender
- DagsHub
- Gotit.pub
- Hugging Face
- Latent Inference-Time Guidance of Time Series Foundation Models
- Loss-Guided Pretraining Data Selection for Time-Series Foundation Models
- RACE: Residual-Aware Test-Time Adaptation for Neighbor-Rich Time-Series Foundation Model Forecasting
- ScienceCast
- Time Series Foundation Models
- TSFMs
- When, Not How Much: Evaluating Time-Series Foundation Models on Sparse Events
AI-generated summary · Google Gemini · from 4 sources. How we write summaries →
