Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
AgentGA optimizes autonomous code-generation by evolving the "agent seed" – the initial task prompt and parent archives that initialize the agent's workspace. This outer-loop optimization searches over reusable starting conditions instead of directly editing code, enabling inheritance of artifacts across generations. Experiments on the Weco-Kaggle Lite benchmark show AgentGA achieves 74.52% Exceeds % of Human, significantly outperforming the AIDE baseline (54.15%), demonstrating the effectiveness of agent-seed optimization.
Forget directly editing code – AgentGA evolves entire autonomous code-generation runs by optimizing the agent's initial conditions, leading to significant performance gains.
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 from a reset 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. On the 10 benchmark runs reported here, AgentGA averages 74.52% Exceeds % of Human versus 54.15% for AIDE. Across 1135 parent-child comparisons, descendants given parent archives outperform runs started from scratch, indicating that inherited artifacts improve later autonomous runs. These findings support agent-seed optimization as a practical design point for autonomous code-search systems.