AIM: Agentic Idea Management for Automated Research
Hyeong Kyu Choi, Bhavana Dalvi Mishra, Chun-Liang Li and colleagues at Google Cloud AI Research and UW-Madison introduce AIM (Agentic Idea Manager), which organizes automated research around explicit research ideas instead of directly editing solution code.
Ask this paper
Two kinds of research agent. The paper separates solution-driven search (AIDE, AlphaEvolve, EvoX edit code directly) from idea-driven search (AI Scientist-v2, ScientistOne reason over ideas and delegate implementation), and names three problems for the latter: idea organization, idea selection and idea-solution integrity.
Bayesian-optimization structure. An Agentic Surrogate clusters ideas across lineages and ranks them by experimental evidence; an Agentic Acquisition step balances exploration and exploitation at both cluster and idea level.
Auditing and budget. A Solution Auditor checks that each implementation matches its idea and that the task was not gamed, and a Resource Planner adjusts parallelism under a fixed experimental budget.
Results. On 10 AutoLab tasks AIM averages 67.0% on System Optimization and 55.8% on Model Development & CUDA, 1.6 and 4.9 points above ScientistOne, and reaches ScientistOne's best score up to 3.1x faster in wall-clock time.
Theory. Explicit idea-level allocation makes coverage of distinct directions controllable, which matters most when only a small fraction of plausible directions are competitive.
Abstract
Frontier LLMs are increasingly used to automate scientific research through iterative search. We distinguish idea-driven search from solution-driven search and identify three core challenges: organizing evolving research ideas, selecting promising directions, and maintaining alignment between ideas and their implementations. To address these challenges, we introduce the Agentic Idea Manager (AIM), a fully autonomous framework for managing and exploring research directions in idea-driven automated research. Inspired by Bayesian optimization, AIM uses an Agentic Surrogate and an Agentic Acquisition mechanism to organize discovered ideas and guide their selection. A Solution Auditor maintains idea-solution integrity, while a Resource Planner adaptively allocates the remaining experimental budget across parallel search branches. Experiments on 10 AutoLab benchmark tasks show that AIM surpasses the strongest baseline by 1.6 percentage points on System Optimization tasks and 4.9 percentage points on long-horizon Model Development & CUDA tasks. Notably, AIM reaches the best baseline performance up to 3.1x faster in wall-clock time. We further provide a theoretical analysis of when searching over ideas becomes beneficial. Our analysis shows that explicit idea-level allocation makes semantic coverage directly controllable, and that broader coverage becomes increasingly valuable when competitive research directions are sparse among many plausible alternatives. Project Page: https://imhgchoi.github.io/agentic-idea-manager/