ML-Master: Towards AI-for-AI via Integration of Exploration and Reasoning

2025-06-19

AI Summary

Existing LLM-driven AI-for-AI agents struggle to reuse experience accumulated during exploration, resulting in low efficiency and suboptimal performance. Method: This paper proposes a novel AI4AI paradigm featuring dynamic synergy between exploration and reasoning, centered on a selective scope memory mechanism. By introducing a tunable memory scope, it enables cross-trajectory knowledge distillation and analytical reasoning guidance within parallel multi-trajectory exploration, mitigating context overload. The method integrates task-adaptive context compression, selective memory retrieval, and a reasoning-driven self-optimization architecture. Contribution/Results: Evaluated on MLE-Bench, our approach achieves a mean medal rate of 29.3%, significantly outperforming baselines on medium-complexity tasks. It completes the entire evaluation in just 12 hours—50% of the baseline time budget—demonstrating substantial gains in both effectiveness and efficiency.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Multiagent Systems: Adversarial Agents
• Search and Optimization: Learning to Search

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation

Abstract

As AI capabilities advance toward and potentially beyond human-level performance, a natural transition emerges where AI-driven development becomes more efficient than human-centric approaches. A promising pathway toward this transition lies in AI-for-AI (AI4AI), which leverages AI techniques to automate and optimize the design, training, and deployment of AI systems themselves. While LLM-based agents have shown the potential to realize AI4AI, they are often unable to fully leverage the experience accumulated by agents during the exploration of solutions in the reasoning process, leading to inefficiencies and suboptimal performance. To address this limitation, we propose ML-Master, a novel AI4AI agent that seamlessly integrates exploration and reasoning by employing a selectively scoped memory mechanism. This approach allows ML-Master to efficiently combine diverse insights from parallel solution trajectories with analytical reasoning, guiding further exploration without overwhelming the agent with excessive context. We evaluate ML-Master on the MLE-Bench, where it achieves a 29.3% average medal rate, significantly surpassing existing methods, particularly in medium-complexity tasks, while accomplishing this superior performance within a strict 12-hour time constraint-half the 24-hour limit used by previous baselines. These results demonstrate ML-Master's potential as a powerful tool for advancing AI4AI.

Problem

• Optimizing AI system design via AI-driven automation
• Enhancing exploration-reasoning integration in AI4AI agents
• Improving efficiency in AI solution discovery processes

Innovation

• Integrates exploration and reasoning seamlessly
• Uses selectively scoped memory mechanism
• Achieves superior performance in strict time