TL;DR of the paper, generated automatically.

New paper: EvoTeam, a way for LLM agent teams to improve over time.

• What changes: team setup, interaction rules and agent skills evolve from past runs.
• Team library: keeps reusable teams, selecting or creating one for each job.
• Testing: experiments on four benchmarks, code on GitHub.