MAEBE: Multi-Agent Emergent Behavior Framework
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2025
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866912474131857408 |
|---|---|
| author | Erisken, Sinem Gothard, Timothy Leitgab, Martin Potham, Ram |
| author_facet | Erisken, Sinem Gothard, Timothy Leitgab, Martin Potham, Ram |
| contents | Traditional AI safety evaluations on isolated LLMs are insufficient as multi-agent AI ensembles become prevalent, introducing novel emergent risks. This paper introduces the Multi-Agent Emergent Behavior Evaluation (MAEBE) framework to systematically assess such risks. Using MAEBE with the Greatest Good Benchmark (and a novel double-inversion question technique), we demonstrate that: (1) LLM moral preferences, particularly for Instrumental Harm, are surprisingly brittle and shift significantly with question framing, both in single agents and ensembles. (2) The moral reasoning of LLM ensembles is not directly predictable from isolated agent behavior due to emergent group dynamics. (3) Specifically, ensembles exhibit phenomena like peer pressure influencing convergence, even when guided by a supervisor, highlighting distinct safety and alignment challenges. Our findings underscore the necessity of evaluating AI systems in their interactive, multi-agent contexts. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2506_03053 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | MAEBE: Multi-Agent Emergent Behavior Framework Erisken, Sinem Gothard, Timothy Leitgab, Martin Potham, Ram Multiagent Systems Artificial Intelligence Computation and Language Computers and Society Machine Learning Traditional AI safety evaluations on isolated LLMs are insufficient as multi-agent AI ensembles become prevalent, introducing novel emergent risks. This paper introduces the Multi-Agent Emergent Behavior Evaluation (MAEBE) framework to systematically assess such risks. Using MAEBE with the Greatest Good Benchmark (and a novel double-inversion question technique), we demonstrate that: (1) LLM moral preferences, particularly for Instrumental Harm, are surprisingly brittle and shift significantly with question framing, both in single agents and ensembles. (2) The moral reasoning of LLM ensembles is not directly predictable from isolated agent behavior due to emergent group dynamics. (3) Specifically, ensembles exhibit phenomena like peer pressure influencing convergence, even when guided by a supervisor, highlighting distinct safety and alignment challenges. Our findings underscore the necessity of evaluating AI systems in their interactive, multi-agent contexts. |
| title | MAEBE: Multi-Agent Emergent Behavior Framework |
| topic | Multiagent Systems Artificial Intelligence Computation and Language Computers and Society Machine Learning |
| url | https://arxiv.org/abs/2506.03053 |