MLExpert: Towards a Generalist Agentic Model for Autonomous Machine Learning Engineering
arXiv cs.AI · 16 Oct 2026, 03:56 AEST
MLExpert is a new agentic model designed for autonomous machine learning engineering, enhancing AI development through advanced experimentation techniques. It demonstrates significant performance improvements over previous models, achieving a medal rate increase from 37.9% to 72.7%.
Key facts
- Large language models (LLMs) are expanding from assisting scientific research and engineering to autonomously developing and improving AI systems, advancing AI for AI (AI4AI) as a pathway toward recursive self-improvement (RSI).
- We propose a post-training method for autonomous long-horizon MLE optimization and apply it to train MLExpert, a 35B Mixture-of-Experts agentic model.
- On MLE-Bench Lite, MLExpert improves the average medal rate of the base model from 37.9% to 72.7%, achieving performance comparable to larger frontier models such as Kimi-K3 and GLM-5.3.
Read aloud
MLExpert is a new agentic model designed for autonomous machine learning engineering, enhancing A I development through advanced experimentation techniques. It demonstrates significant performance improvements over previous models, achieving a medal rate increase from thirty seven point nine percent to seventy two point seven percent. Large language models L L M s are expanding from assisting scientific research and engineering to autonomously developing and improving A I systems, advancing A I for A I AI4AI as a pathway toward recursive self-improvement RSI. We propose a post-training method for autonomous long-horizon MLE optimization and apply it to train MLExpert, a 35B Mixture-of-Experts agentic model. On MLE-Bench Lite, MLExpert improves the average medal rate of the base model from thirty seven point nine percent to seventy two point seven percent, achieving performance comparable to larger frontier models such as Kimi-K3 and GLM-five point three.
Sources in this cluster
- MLExpert: Towards a Generalist Agentic Model for Autonomous Machine Learning Engineering
arXiv cs.AI · 16 Oct 2026, 03:56 AEST