TimeRSI: Towards Self-Improved Agentic Deployment for Time-Series Foundation Models
TimeRSI uses coding agents to iteratively improve frozen time-series pipelines and leads 198 forecasting tasks.
TimeRSI is a self-improving agentic deployment system that wraps frozen time-series foundation models. A coding agent turns task descriptions into executable pipeline changes, tests them on historical windows, and uses forecast errors to guide later revisions. An adaptive selector then picks among pipelines using completed forecasts under similar conditions. On 198 TIME and fev-bench tasks, it outperforms all submissions on six point and probabilistic metrics, and the gains hold across coding agents and forecasting models.
- A coding agent revises frozen-model forecasting pipelines from task text and errors.
- An adaptive selector chooses pipelines using completed forecasts under similar conditions.
- TimeRSI beats all submissions on six metrics across 198 TIME and fev-bench tasks.
- Feedback-guided revision outperforms non-iterative pipeline search in controlled comparisons.
From the original article
1 min read at huggingface.co
Time-series foundation models provide strong zero-shot forecast, but deployment for real-world applications still requires task-specific designs for model inputs and outputs. Direct inference can leave domain knowledge and recurring forecast errors unaddressed, while tailored pipelines require substantial human expertise and effort.
By Haoxin Liu, Deqing Fu, Lucheng Fu · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.