Communication Dans Un Congrès Année : 2025

Using (Traditional) Multi-Agent principles to suggest improvements for Agentic AI platforms

Résumé

Multi-Agent Systems (MAS) have been a key AI subfield since the 1980s, with strong academic foundations in intelligent agents and agent-oriented programming. The rise of LLMs in the 2020s has led to "AI Agents" with greater autonomy, renewing interest in MAS and inspiring new platforms. While traditional MAS researchers are starting to integrate LLMs, the emerging, industry-driven Agentic AI community often overlooks MAS formalisms, leading to narrower, less rigorous definitions. This paper uses MAS formalism to clarify Agentic AI concepts, highlight current limitations, and suggest paths for improvement.

Fichier principal
Vignette du fichier
article_chloeEDF_EUMAS25_sub1.pdf (115.25 Ko) Télécharger le fichier
EUMAS_Agent_Toolkits_2025_Community_Session_Report.pdf (409.97 Ko) Télécharger le fichier

Dates et versions

hal-05176616 , version 1 (22-07-2025)

Licence

Identifiants

  • HAL Id : hal-05176616 , version 1

Citer

Guillaume Muller, Somsakun Maneerat, Heloise Pajot, Chloé Petridis, Richard Niamke. Using (Traditional) Multi-Agent principles to suggest improvements for Agentic AI platforms. Agent Toolkits Community Session : The 22nd European Conference on Multi-Agent Systems - EUMAS 2025, Sep 2025, Bucharest, Romania. ⟨hal-05176616⟩
936 Consultations
555 Téléchargements

Partager

  • More