Gespeichert in:
| Hauptverfasser: | , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2026
|
| Schlagworte: | |
| Online-Zugang: | https://arxiv.org/abs/2604.14655 |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866910206933336064 |
|---|---|
| author | Tan, David Y. Y. Chin, Kellie Zhang, Jingxian |
| author_facet | Tan, David Y. Y. Chin, Kellie Zhang, Jingxian |
| contents | We present AgentGA, a framework that evolves autonomous code-generation runs by optimizing the agent seed: the task prompt plus optional parent archives that initialize a fresh workspace. The outer loop searches over these reusable starting conditions rather than editing code directly. Each generation launches a fresh autonomous run in an isolated workspace, while selected parent archives provide inherited artifacts that descendants can inspect and reuse. AgentGA couples a population-level genetic algorithm with long-horizon agents; selection uses deterministic 1:1 elite tournaments and operator allocation is adapted online with a modified Hedge controller. We instantiate the approach for tabular AutoML on the 16-competition Weco-Kaggle Lite benchmark. Across the full benchmark, AgentGA averages 71.90% Exceeds % of Human versus 51.38% for the AIDE reference, winning 15/16 competitions. Within AgentGA runs, descendants conditioned on inherited parent archives win 51.9% of 1,680 parent-child tournaments versus 8.6% for de novo proposals. These results support agent-seed optimization as a practical design choice for autonomous code-search systems. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2604_14655 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | AgentGA: Evolving Code Solutions in Agent-Seed Space Tan, David Y. Y. Chin, Kellie Zhang, Jingxian Artificial Intelligence Machine Learning We present AgentGA, a framework that evolves autonomous code-generation runs by optimizing the agent seed: the task prompt plus optional parent archives that initialize a fresh workspace. The outer loop searches over these reusable starting conditions rather than editing code directly. Each generation launches a fresh autonomous run in an isolated workspace, while selected parent archives provide inherited artifacts that descendants can inspect and reuse. AgentGA couples a population-level genetic algorithm with long-horizon agents; selection uses deterministic 1:1 elite tournaments and operator allocation is adapted online with a modified Hedge controller. We instantiate the approach for tabular AutoML on the 16-competition Weco-Kaggle Lite benchmark. Across the full benchmark, AgentGA averages 71.90% Exceeds % of Human versus 51.38% for the AIDE reference, winning 15/16 competitions. Within AgentGA runs, descendants conditioned on inherited parent archives win 51.9% of 1,680 parent-child tournaments versus 8.6% for de novo proposals. These results support agent-seed optimization as a practical design choice for autonomous code-search systems. |
| title | AgentGA: Evolving Code Solutions in Agent-Seed Space |
| topic | Artificial Intelligence Machine Learning |
| url | https://arxiv.org/abs/2604.14655 |