AgentGA: Evolving Code Solutions in Agent-Seed Space

2026-04-16 · Source: Machine Learning · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering, Data Science & Analytics · Depth: Expert, quick

Summary

AgentGA is a novel framework designed to evolve autonomous code-generation runs by optimizing the "agent seed," which comprises the task prompt and optional parent archives used to initialize a workspace. This system employs an outer loop that searches over these reusable starting conditions, rather than directly modifying code. Each generation initiates a new autonomous run from a clean workspace, with selected parent archives providing inherited artifacts for inspection and reuse by descendants. AgentGA integrates a population-level genetic algorithm with long-horizon agents, utilizing deterministic 1:1 elite tournaments for selection and an online-adapted Hedge controller for operator allocation. When instantiated for tabular AutoML on the 16-competition Weco-Kaggle Lite benchmark, AgentGA achieved an average of 74.52% Exceeds % of Human across 10 benchmark runs, outperforming AIDE's 54.15%. Furthermore, 1135 parent-child comparisons demonstrated that inherited artifacts significantly enhance subsequent autonomous runs.

Key takeaway

For research scientists developing autonomous code-generation systems, AgentGA demonstrates that optimizing the "agent seed" and enabling artifact inheritance significantly boosts performance. You should consider integrating genetic algorithms with agent-seed optimization and mechanisms for passing inherited artifacts to improve the efficacy and efficiency of your code-search frameworks, as this approach outperformed AIDE on tabular AutoML benchmarks.

Key insights

AgentGA optimizes autonomous code generation by evolving agent seeds, improving performance through inherited artifacts.

Principles

• Optimize agent seeds, not code directly.
• Inherited artifacts improve autonomous runs.

Method

AgentGA uses a genetic algorithm with long-horizon agents, 1:1 elite tournaments, and an online Hedge controller to evolve agent seeds for code generation.

In practice

• Use agent-seed optimization for code search.
• Implement artifact inheritance in agent systems.

Topics

• AgentGA Framework
• Agent-Seed Optimization
• Autonomous Code Generation
• Genetic Algorithms
• Tabular AutoML

Best for: Research Scientist, AI Scientist, Machine Learning Engineer

Related on AIssential

• What Do Evolutionary Coding Agents Evolve? — Evolutionary coding agents often re-introduce deleted code and achieve gains through hyperparameter tuning, not just novel algorithmic structures.
• New agent framework matches human-engineered AI systems — and adds zero inference cost to deploy — Group-Evolving Agents (GEA) enable AI systems to autonomously improve by fostering collective learning and shared innovation among agent groups.
• AlphaEvolve: How our Gemini-powered coding agent is scaling impact across fields — AlphaEvolve, a Gemini-powered agent, autonomously discovers and optimizes algorithms, driving significant advancements across science, infrastructure, and commercial applications.
• Will AI Agents Replace Software Developers? | Front Page — EVO automates code optimization through AI agents, iterative experimentation, and learning from both successes and failures.
• Advent of Slop: A Guest Post by Claude — AI agents can autonomously solve complex coding challenges, including optimization and input generation.
• AI-PROPELLER: Warehouse-Scale Interprocedural Code Layout Optimization with AlphaEvolve — AI-PROPELLER uses an evolutionary agentic workflow and hardware-based feedback to achieve significant interprocedural code layout optimization for large-scale applications.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Machine Learning.