New SELA framework uses VLM agents for explainable time series event detection

Researchers have developed a new framework called SELA for detecting events in multivariate time series data using neuro-symbolic VLM agents. This approach, termed Language-guided TSED, leverages textual descriptions of events to identify corresponding intervals in signals with minimal or no labeled data. The system utilizes an Event Logic Tree (ELT) knowledge representation to translate linguistic descriptions into structured temporal logic, enabling the grounding of signal primitives and the generation of faithful, tree-structured explanations for detected events. Experiments on real-world energy and climate datasets demonstrate SELA's improvement over existing supervised and zero/few-shot time series reasoning baselines. AI

IMPACT Introduces a novel neuro-symbolic approach for explainable event detection in time series, potentially improving applications in critical domains.

RANK_REASON The cluster contains an academic paper detailing a new method for time series event detection. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• Event Logic Tree
• Language-guided TSED
• SELA
• Time Series Event Detection
• VLM agents

• paper
• product

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