https://arxiv.org/pdf/2308.00352
Solutions to more complex tasks, however, are complicated through logic inconsistencies due to cascading hallucinations caused by naively chaining LLMs
기존의 multi-agent의 할루시네이션으로 인한 LLM을 연결할 때의 생기는 inconsistency 문제를 해소하기 위한 metagpt 프레임워크
아래의 SOP를 prompt로 바꿔 streamlined workflow 제공
thus allowing agents with human-like domain expertise to verify intermediate results and reduce errors
이를 통해 인간과 같은 도메인 전문성을 가진 에이전트들이 중간 결과를 검증하고 오류를 줄일 수 있게 한다
SOP란?
SOP는 특정 작업이나 프로세스를 일관되고 효율적으로 수행하기 위해 개발된 상세한 문서화된 지침
These SOPs play a critical role in supporting task decomposition and effective coordination
all employees follow a strict and streamlined workflow, and all their handovers must comply with certain established standards. This reduces the risk of hallucinations caused by idle chatter between LLMs, particularly in role-playing frameworks, like: “Hi, hello and how are you?” – Alice (Product Manager); “Great! Have you had lunch?” – Bob (Architect).
HumanEval, MBPP 벤치마크 사용
: An example of the communication protocol (left) and iterative programming with executable feedback (right)
