TimeRSI: Towards Self-Improved Agentic Deployment for Time-Series Foundation Models
Oct 14, 2026·,,
,,,·
1 min read
Haoxin Liu
Deqing Fu
Lucheng Fu
Yiqiao Jin
Rushi Qiang
Harshavardhan Kamarthi
B. Aditya Prakash
Abstract
TimeRSI develops task-specific forecasting pipelines around frozen time-series foundation models. A coding agent translates task descriptions into executable changes, evaluates revisions on historical forecasting windows, and uses forecast errors to guide further improvements. At deployment, an adaptive selector chooses among the resulting pipelines using feedback from completed forecasts under similar conditions.
Type
Publication
arXiv preprint arXiv:2610.19725
Overview
TimeRSI develops task-specific forecasting pipelines around frozen time-series foundation models. A coding agent translates task descriptions into executable changes, evaluates revisions on historical forecasting windows, and uses forecast errors to guide further improvements. At deployment, an adaptive selector chooses among the resulting pipelines using feedback from completed forecasts under similar conditions.
