Artificial intelligence(AI)agents driven by large language models have demonstrated substantial advancements in addressing complex problem-solving tasks.While current AI agents’collaborations are predominantly organi...Artificial intelligence(AI)agents driven by large language models have demonstrated substantial advancements in addressing complex problem-solving tasks.While current AI agents’collaborations are predominantly organized by centralized platforms,this organizational model inherently introduces systemic challenges such as data monopolies and interoperability barriers,thereby hindering cross-domain collaborative potential.Although current multi-agent collaboration methods have enabled secure data and resource sharing,they still fall short in utilizing distributed knowledge and adopting validatable collaborative workflows.To overcome these limitations,this paper proposes a distributed multi-agent collaboration mechanism leveraging the smart-contracts-based validation method with specifically designed planning,execution and validation smart contracts for robust workflow validation.An illustrative example of decentralized science collaborative writing is provided to demonstrate the applicability of the proposed mechanism.To further bridge the data silos and autonomously update the smart contracts,we employ decentralized autonomous organizations and operations(DAOs)and adopt an auction-based tool supplier matching model to architect the organizing and incentive scheme to enhance selfadaptivity and scalability.To further evaluate the performance of the proposed mechanism,we delineate a hierarchical indicator system and discuss the corresponding key technologies.This paper outlines a smart-contract-based approach to enhance the reliability of multi-agent collaboration,which could serve as a methodological basis for constructing future distributed multiagent collaboration ecosystems.展开更多
基金supported by the Science and Technology Development Fund,Macao SAR(Nos.0093/2023/RIA2,0157/2024/RIA2,and 0145/2023/RIA3).
文摘Artificial intelligence(AI)agents driven by large language models have demonstrated substantial advancements in addressing complex problem-solving tasks.While current AI agents’collaborations are predominantly organized by centralized platforms,this organizational model inherently introduces systemic challenges such as data monopolies and interoperability barriers,thereby hindering cross-domain collaborative potential.Although current multi-agent collaboration methods have enabled secure data and resource sharing,they still fall short in utilizing distributed knowledge and adopting validatable collaborative workflows.To overcome these limitations,this paper proposes a distributed multi-agent collaboration mechanism leveraging the smart-contracts-based validation method with specifically designed planning,execution and validation smart contracts for robust workflow validation.An illustrative example of decentralized science collaborative writing is provided to demonstrate the applicability of the proposed mechanism.To further bridge the data silos and autonomously update the smart contracts,we employ decentralized autonomous organizations and operations(DAOs)and adopt an auction-based tool supplier matching model to architect the organizing and incentive scheme to enhance selfadaptivity and scalability.To further evaluate the performance of the proposed mechanism,we delineate a hierarchical indicator system and discuss the corresponding key technologies.This paper outlines a smart-contract-based approach to enhance the reliability of multi-agent collaboration,which could serve as a methodological basis for constructing future distributed multiagent collaboration ecosystems.
