FreeEvolve: Learning to Evolve Beyond Fixed Loops
Agent evolvers automate the design of the prompts, skills and workflows around language model agents, yet the optimization process they follow is still designed by hand: a fixed search loop decides how candidates are evaluated, which are kept and when the search stops.
Key points
- We propose FREEEVOLVE, which automates this process as well.
- These decisions follow an editable evolution skill, which we improve through meta-evolution by scoring each candidate skill on the fresh target agent it produces.
- On tau3-bench, ARC-AGI-2, ARC-AGI-3 and Terminal-Bench 2.1, FREEEVOLVE controls the evolution campaign by itself, yet improves the primary held-out metric by 13.6 points on average and matches or exceeds hand-designed evolvers.
- The learned process keeps improving with experience: meta-evolved skills add 6.9 points over the seed skill on fresh target agents, demonstrating transferability across environments.
Sources (1)
- [1]FreeEvolve: Learning to Evolve Beyond Fixed LoopsarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 6, 10:47 PM
Agent evolvers automate the design of the prompts, skills and workflows around language model agents, yet the optimization process they follow is still designed by hand: a fixed search loop decides how candidates are evaluated, which are kept and when the search stops.
We propose FREEEVOLVE, which automates this process as well.
Extractive summary: sentences quoted from the sources.