ConceptTS framework uses LLMs for interpretable time-series forecasting

Researchers have developed ConceptTS, a new framework for interpretable time-series forecasting. This system leverages large language models to identify and define human-readable concepts relevant to the forecasting task. These concepts are then integrated into the model through three distinct bottlenecks, allowing for a transparent decision-making process and enabling direct concept-level interventions. Experiments on air quality data demonstrated that ConceptTS achieves competitive accuracy with existing black-box models while providing meaningful concept activations. AI

IMPACT Introduces a novel method for making complex time-series models more transparent and understandable.

RANK_REASON The cluster contains a research paper detailing a new interpretable forecasting framework. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• alphaXiv
• arXiv
• Beijing Multi-Site Air Quality dataset
• CatalyzeX
• ConceptTS
• Connected Papers
• DagsHub
• Gotit.pub
• Hugging Face
• IArxiv
• Litmaps
• ScienceCast
• scite Smart Citations

• paper
• product

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →