Researchers present AIDE^2, a system that implements a loop of recursive self-improvement for a frontier AI research agent by optimizing its own code. The system proposes changes to its internal logic, benchmarks modified versions on AI R&D tasks, and retains only those updates that perform best on hidden evaluations.
- In an autonomous 8-day run, AIDE^2 discovered seven successive improvements, including new search policies and memory mechanisms for context compression.
- These gains generalized to four held-out benchmarks spanning machine learning engineering, heuristic algorithm engineering, and physics-based weather forecasting.
- On all four benchmarks, the strongest discovered agent matched or exceeded a human-engineered production research agent ranked among the strongest on FML-Bench.
- On a separate held-out task family, the discovered agents exhibited reduced reward hacking, with the rate falling from 55% to 32%, seven percentage points below the human-engineered agent.
These results demonstrate that an AI research agent can improve its own research efficiency through recursive self-improvement and that these gains transfer to tasks and domains the loop never encountered.