MLExpert: Towards a Generalist Agentic Model for Autonomous Machine Learning Engineering
Shangheng Du, Jinxin Shi, Shuaiyu Zhang and colleagues (Shanghai AI Laboratory and East China Normal University) post-train a 35B MoE model for long-horizon machine learning engineering and nearly double its MLE-Bench Lite medal rate.
Ask this paper
Task factory. Kaggle-MLE Factory builds executable tasks, and an interaction harness lets the model run and organize its own experiments.
Training. Hierarchical agentic SFT transfers general tool-use and coding skill to MLE through outcome-guided supervision. Windowed Interaction RL (WIRL) treats the decisions between two experiment results as one window with a shared outcome reward, and asynchronous collection lets updates start before a trajectory ends.
Result. The average medal rate on MLE-Bench Lite rises from 37.9% for the base model to 72.7%, comparable to Kimi-K3 and GLM-5.3.
Transfer. Gains carry over to NatureBench Lite, agent scaffold optimization and algorithm optimization. Code and weights are released.
Abstract
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). Workflow-driven machine learning engineering (MLE) systems provide a practical path toward this pursuit through complex orchestration that manages long-horizon experimentation, while recent MLE systems adopt model-directed interaction, allowing models to autonomously organize experiments and adapt optimization strategies based on execution feedback. However, feedback from each experiment reflects several preceding decisions, and a long MLE run contains many such costly experiments. Training must therefore select useful supervision from long interaction trajectories and learn from intermediate outcomes. 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. We design the Kaggle-MLE Factory to construct executable tasks and an interaction harness for model-directed experimentation. Hierarchical agentic SFT transfers general tool-use and coding capabilities to MLE through outcome-guided supervision. Windowed Interaction RL (WIRL) treats the multi-turn decisions between successive experiment results as an execution window and assigns it a shared outcome-based reward. Asynchronous collection and policy correction support policy updates before the trajectory ends. 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. Further experiments show that MLExpert improves performance on NatureBench Lite, agent scaffold optimization, and algorithm optimization, suggesting that the learned optimization capability generalizes to other tasks.