Editing Recursive self-improvement (section)

== Experimental research ==
In 2023, the Voyager agent learned to accomplish diverse tasks in [[Minecraft]] by iteratively prompting a LLM for code, refining this code based on feedback from the game, and storing the programs that work in an expanding skills library.<ref>{{Cite web |last=Schreiner |first=Maximilian |date=2023-05-28 |title=Minecraft bot Voyager programs itself using GPT-4 |url=https://the-decoder.com/minecraft-bot-voyager-programs-itself-using-gpt-4/ |access-date=2025-05-20 |website=The decoder |language=en-US}}</ref>

In 2024, researchers proposed the framework "STOP" (Self-optimization Through Program Optimization), in which a "scaffolding" program recursively improves itself using a fixed LLM.<ref>{{Cite journal |date=2024 |title=Self-Taught Optimizer (STOP): Recursively Self-Improving Code Generation |url=https://arxiv.org/pdf/2310.02304 |journal=COLM conference}}</ref>

[[Meta AI]] has performed various research on the development of large language models capable of self-improvement. This includes their work on "Self-Rewarding Language Models" that studies how to achieve super-human agents that can receive super-human feedback in its training processes.<ref>{{cite arXiv |eprint=2401.10020 |class=cs.CL |first1=Weizhe |last1=Yuan |first2=Richard Yuanzhe |last2=Pang |title=Self-Rewarding Language Models |date=2024-01-18 |last3=Cho |first3=Kyunghyun |last4=Sukhbaatar |first4=Sainbayar |last5=Xu |first5=Jing |last6=Weston |first6=Jason}}</ref>

In May 2025, Google DeepMind unveiled [[AlphaEvolve]], an [[Evolutionary computation|evolutionary]] coding agent that uses a LLM to design and optimize algorithms. Starting with an initial algorithm and performance metrics, AlphaEvolve repeatedly mutates or combines existing algorithms using a LLM to generate new candidates, selecting the most promising candidates for further iterations. AlphaEvolve has made several algorithmic discoveries and could be used to optimize components of itself, but a key limitation is the need for automated evaluation functions.<ref>{{Cite web |last=Tardif |first=Antoine |date=2025-05-17 |title=AlphaEvolve: Google DeepMind’s Groundbreaking Step Toward AGI |url=https://www.unite.ai/alphaevolve-google-deepminds-groundbreaking-step-toward-agi/ |access-date=2025-05-20 |website=Unite.AI |language=en-US}}</ref>