Uloženo v:
Podrobná bibliografie
Hlavní autor: Gomes, Renato Aparecido
Médium: Recurso digital
Jazyk:
Vydáno: Zenodo 2026
Témata:
On-line přístup:https://doi.org/10.5281/zenodo.19325749
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
_version_ 1866901355367497728
author Gomes, Renato Aparecido
author_facet Gomes, Renato Aparecido
contents Current multi-agent systems built on large language models rely on static orchestration patterns that fail to adapt when task demands shift dynamically. We present SYMBIONT (Symbiotic Multi-pattern Bio-intelligent Organism for Networked Tasks), a framework integrating eight biological swarm patterns into a unified organism for LLM agent coordination: mycorrhizal fungus networks (adaptive routing), slime mold topology optimization, ant caste polymorphism (5-caste agent specialization), honeybee waggle dance (collective decisions), termite mound architecture (stigmergy and homeostasis), starling murmuration (real-time reflexes), wolf/mole-rat governance (contextual leadership), and dolphin pod dynamics (ephemeral coalitions). The key design principle is functional inspiration, not structural replication. SYMBIONT is implemented in approximately 3,700 lines of Python with five agent castes mapping to LLM cost tiers, nine emergent laws, and a five-phase task execution lifecycle. Submitted to ANTS 2026 (15th International Conference on Swarm Intelligence, Darmstadt, Germany).
format Recurso digital
id zenodo_https___doi_org_10_5281_zenodo_19325749
institution Zenodo
language
publishDate 2026
publisher Zenodo
record_format zenodo
spellingShingle SYMBIONT: Unifying Eight Biological Swarm Patterns for LLM Agent Coordination
Gomes, Renato Aparecido
swarm intelligence
multi-agent systems
bio-inspired computing
LLM agents
collective intelligence
self-organization
SYMBIONT
Current multi-agent systems built on large language models rely on static orchestration patterns that fail to adapt when task demands shift dynamically. We present SYMBIONT (Symbiotic Multi-pattern Bio-intelligent Organism for Networked Tasks), a framework integrating eight biological swarm patterns into a unified organism for LLM agent coordination: mycorrhizal fungus networks (adaptive routing), slime mold topology optimization, ant caste polymorphism (5-caste agent specialization), honeybee waggle dance (collective decisions), termite mound architecture (stigmergy and homeostasis), starling murmuration (real-time reflexes), wolf/mole-rat governance (contextual leadership), and dolphin pod dynamics (ephemeral coalitions). The key design principle is functional inspiration, not structural replication. SYMBIONT is implemented in approximately 3,700 lines of Python with five agent castes mapping to LLM cost tiers, nine emergent laws, and a five-phase task execution lifecycle. Submitted to ANTS 2026 (15th International Conference on Swarm Intelligence, Darmstadt, Germany).
title SYMBIONT: Unifying Eight Biological Swarm Patterns for LLM Agent Coordination
topic swarm intelligence
multi-agent systems
bio-inspired computing
LLM agents
collective intelligence
self-organization
SYMBIONT
url https://doi.org/10.5281/zenodo.19325749