MLExpert: Towards a Generalist Agentic Model for Autonomous Machine Learning Engineering

Hugging Face Daily Papers Papers
agentic-ai machine-learning-engineering reinforcement-learning mixture-of-experts ai4ai post-training

Summary

提出了 MLExpert，一个 35B 混合专家（MoE）智能体模型，通过分层监督微调和窗口化交互强化学习（WIRL）自主优化长周期机器学习工程工作流，在 MLE-Bench Lite 上将基模型奖牌率从 37.9% 提升至 72.7%，媲美更大的前沿模型。

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.
Original Article
View Cached Full Text

Cached at: 10/16/26, 07:03 AM

Paper page - MLExpert: Towards a Generalist Agentic Model for Autonomous Machine Learning Engineering

Source: https://huggingface.co/papers/2610.22344

Abstract

A post-training framework using hierarchical supervised fine-tuning and windowed interaction reinforcement learning trains a 35B mixture-of-experts agent to autonomously optimize long-horizon machine learning engineering workflows.

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 towardrecursive self-improvement(RSI). Workflow-drivenmachine learning engineering(MLE) systems provide a practical path toward this pursuit through complex orchestration that manages long-horizon experimentation, while recent MLE systems adoptmodel-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 35BMixture-of-Expertsagentic model. We design the Kaggle-MLE Factory to construct executable tasks and an interaction harness for model-directed experimentation.Hierarchical agentic SFTtransfers 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 collectionandpolicy correctionsupport policy updates before the trajectory ends. OnMLE-BenchLite, 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.

View arXiv pageView PDFGitHub4Add to collection

Get this paper in your agent:

hf papers read 2610\.22344

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper1

#### InternScience/MLExpert Text Generation• 35B• Updatedabout 1 hour ago • 7 • 6

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2610.22344 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2610.22344 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

Explaining the Roles of Experts in Mixture-of-Experts LLMs for Personalization

Hugging Face Daily Papers

本文提出对稀疏混合专家（MoE）大语言模型进行个性化，通过训练 crosscoder 学习可解释的个性化隐变量，并利用因果干预验证这些隐变量及其关联专家足以复现个性化行为，相比稠密模型降低了多义性并实现了局部可定位的控制。