Source: AutoGen: Enabling Next-Gen LLM Applications via Multi-Agent Conversation (Wu et al., 2023)
The original AutoGen paper introducing conversable agents and conversation programming for multi-agent LLM apps.
Contents
- What: research paper, arXiv 2308.08155
- Author: Qingyun Wu, Gagan Bansal, Jieyu Zhang, Yiran Wu, Beibin Li and 9 others (Microsoft Research and collaborators)
- Date: v1 2023-08-16, v2 2023-10-03
- Original: https://arxiv.org/abs/2308.08155
Summary
Introduces AutoGen, an open-source framework where applications are built from multiple customisable agents that converse to solve tasks. Developers program interaction patterns in natural language and code; agents can combine LLMs, human input and tools.
Key claims
- Agents are customisable and conversable and can operate in several modes (LLM, human, tool-backed).
- Conversation programming defines flexible interaction patterns in natural language and code.
- Evaluated on maths, coding, QA, operations research, online decision-making and entertainment.