EurekAgent: Agent Environment Engineering is All You Need For Autonomous Scientific Discovery

2026-06-11

AI Summary

This work addresses a critical bottleneck in autonomous scientific discovery, which has shifted from agent workflow design to environment construction. The authors propose a new paradigm—“environment engineering”—that systematically designs execution environments to support open-ended exploration and reliable collaboration among agents. This framework incorporates four key mechanisms: permission control, file system and Git-based coordination, budget-aware exploration, and human-in-the-loop intervention, collectively guiding agents toward beneficial behaviors while suppressing harmful ones. Evaluated across mathematical discovery, kernel engineering, and machine learning tasks, the approach achieves new state-of-the-art results, including establishing a new record for the 26-circle packing problem at an API cost under $11, thereby significantly enhancing the efficiency and reliability of autonomous research agents.

Abstract

LLM-based agents have shown increasing potential in automating scientific discovery. Given an optimizable metric and an execution environment, they can propose, validate, and iterate scientific solutions, and have produced results that outperform human-designed approaches. As model capabilities continue to improve, we argue that the bottleneck for autonomous scientific discovery is shifting from prescribing agent workflows to designing agent environments: the resources, constraints, and interfaces that shape agent behavior. We frame this as environment engineering: building environments that amplify productive behaviors, such as open-ended exploration, systematic artifact management, and inter-agent collaboration, while suppressing harmful behaviors, such as reward hacking and high-friction human oversight. We present EurekAgent, an environment-engineered agent system for metric-driven autonomous scientific discovery. EurekAgent engineers the environment along four dimensions: permissions engineering for bounded agent execution and isolated evaluation; artifact engineering for filesystem and Git-based collaboration; budget engineering for budget-aware exploration; and human-in-the-loop engineering for easy human supervision and intervention. EurekAgent sets new state-of-the-art results on multiple mathematics, kernel engineering, and machine learning tasks, including new state-of-the-art 26-circle packing results discovered with less than $11 in total API cost. We open-source our code and results, and call for environment engineering as a core research direction for developing reliable autonomous research agents.

Problem

• agent environment engineering
• autonomous scientific discovery
• LLM-based agents
• reward hacking
• human-in-the-loop

Innovation

• Environment Engineering
• Autonomous Scientific Discovery
• LLM-based Agents
• Artifact Management
• Human-in-the-Loop