Research map / Multi-agent systems and games / LLM agents
Resilience in Ambient Multi-Agent LLMs via Decentralized Bio-Autonomic Control and Immune-Inspired Anomaly Detection
Nastaran Darabi, Devashri Naik, Sina Tayebati, Dinithi Jayasuriya, Amit Ranjan Trivedi
Read the paperCheck your idea against 137,217 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Proposes a decentralized bio-autonomic control framework with immune-inspired anomaly detection to improve resilience of LLM multi-agent systems in Ambient Intelligence environments.
Problem
Reliability of autonomous LLM agents deployed at scale in Ambient Intelligence environments, where isolated adaptive control, anomaly detection, or trust modeling are fragmented and scenario-specific.
Approach
Unifies adaptive control, anomaly detection, and trust modeling into a decentralized bio-autonomic control architecture for multi-agent LLM systems, drawing inspiration from biological and immune systems.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add the Bio-Autonomic framework on top of the LLM backbone instead of using AutoAgents | multi-agent LLM system · HumanEval | score (0-100): 81.23 -> 88.52 (GPT-4o) | Improved |
| add the Bio-Autonomic framework instead of H-MAS | multi-agent LLM system · CIAR | score (0-100): 62.34 -> 72.89 (GPT-4o) | Improved |
| add the Bio-Autonomic framework to the Collab-Overcooked agent | GPT-4o · Collab-Overcooked Level 6 | SR (%): 4.00 -> 20.00 | Improved |
| add the Bio-Autonomic framework to a small model | Llama3.1-8B · Collab-Overcooked Level 1 | SR (%): 4.00 -> 72.00 | Improved |
| add Bio-Autonomic under AUTOTRANSFORM fault | GPT-4o · Collab-Overcooked Level 1 | SR (%): 65.80 -> 95.00 | Improved |
| use Bio-Autonomic instead of MADDPG under 20% semantic faults | multi-agent LLM system · AUTOINJECT 20% | PDF (%): -65.0 -> -4.8 | Improved |
| use Bio-Autonomic instead of AutoAgents under 20% semantic faults | multi-agent LLM system · AUTOINJECT 20% | ERR (%): 12.5 -> 93.4 | Improved |
| add direct trust ledger (versus none) (ablation) | Bio-Autonomic · AUTOTRANSFORM single malicious agent | CFHR (%): 65.4 -> 15.2 | Improved |
| add gossip-based reputation dissemination to direct trust (ablation) | Bio-Autonomic · AUTOTRANSFORM single malicious agent | MAI: 0.73 -> 0.94 | Improved |
| add gossip-based reputation dissemination to direct trust (ablation) | Bio-Autonomic · AUTOTRANSFORM single malicious agent | ERR (%): 74.0 -> 93.4 | Improved |
Closest papers by meaning
- AgentNet: Decentralized Evolutionary Coordination for LLM-based Multi-Agent Systems (NeurIPS 2025 · similarity 0.87)
- ResMAS: Resilience Optimization in LLM-based Multi-agent Systems (AAAI 2026 · similarity 0.86)
- Rethinking the Reliability of Multi-agent System: A Perspective from Byzantine Fault Tolerance (AAAI 2026 · similarity 0.85)
- On the Resilience of LLM-Based Multi-Agent Collaboration with Faulty Agents (ICML 2025 · similarity 0.85)
- SafeSieve: From Heuristics to Experience in Progressive Pruning for LLM-based Multi-Agent Communication (AAAI 2026 · similarity 0.85)
- ORBIT: A Framework for Multi-Agent Security Evaluations (NeurIPS 2026 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
