Recursive self-improvement of AI research agents
Dhruv Srikanth, Zhengyao Jiang and colleagues at Weco AI present AIDE^2, a loop in which a frontier AI research agent edits its own code, benchmarks the new versions on AI R&D tasks and keeps the edits that score best on hidden evaluations.
Ask this paper
The loop. Each accepted rewrite becomes the agent that the next round edits, so improvements compound across rounds.
Eight-day run. An autonomous run found seven successive improvements, from a new search policy to memory mechanisms that compress and manage the agent's growing context.
Transfer. Gains carry over to four held-out benchmarks in ML engineering, heuristic algorithm engineering and physics-based weather forecasting, the last of which is outside the selection distribution.
Against a human-built agent. On all four, the best discovered agent matches or beats a human-engineered production research agent that ranks among the strongest on FML-Bench.
Less reward hacking. On a separate held-out task family, the reward-hacking rate fell from 55% to 32% during the run, 7 points below the human-engineered agent, although the loop never optimized for it.
Abstract
AI agents are beginning to automate research and development across the AI stack, from improving training efficiency to optimizing inference. A natural next step is to improve the research efficiency of the agents themselves. When an AI research agent's own code is the object of optimization, each accepted rewrite becomes the agent that the next round edits. We refer to this loop as recursive self-improvement. Its significance lies in a long-standing trend, in which increased cumulative spending on R&D yields diminishing returns. Sustained self-improvement offers a way to counter this trend. We present AIDE^2, a system that implements this loop for a frontier AI research agent. It proposes changes to its own code, benchmarks modified versions of itself on a suite of AI R&D tasks, and keeps the changes that perform best on hidden evaluations. In an autonomous 8-day run, AIDE^2 discovered seven successive improvements, ranging from a new search policy to memory mechanisms that compress and manage the agent's growing context. These gains generalize to four held-out benchmarks spanning machine learning engineering, heuristic algorithm engineering, and physics-based weather forecasting, the last of which is out of distribution from the selection tasks. On all four, the strongest discovered agent matches or exceeds a human-engineered production research agent that ranks among the strongest on FML-Bench. On a separate held-out task family, the discovered agents also exhibit reduced reward hacking, a property the loop never explicitly optimized for: the rate falls from 55% to 32% during the run, 7 percentage points below the human-engineered agent. Together, these results show 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.