Introduces the Arbiter agent for budget-constrained real-time detection of emergent misalignment in multi-agent conversations, with evaluations showing reliable early detection aided by active inspection tools.
arXiv preprint arXiv:2506.01080 , year=
5 Pith papers cite this work, alongside 4 external citations. Polarity classification is still indexing.
5
Pith papers citing it
4
external citations · external index
years
2026 5representative citing papers
VLMs preserve linearly separable visual magnitudes and can compare them, yet collapse at symbolic mapping because visual and textual number spaces remain fractured and disjoint.
External evolution beats internal deliberation in collective-action tasks with statistical significance but neither helps in trading, and deliberation never discovers punishment while evolution does.
Adversarial co-evolution of LLM constitutions in public goods games reaches near-parity equilibrium only when fitness is coupled across factions and evaluation uses at least five seeds per generation.
A survey arguing that multi-agent AI systems remain task-centric and lack integrated computational models of human cognition, culture, values, and social cooperation.
citing papers explorer
-
The Arbiter Agent: Continually Monitoring Multi-Agent Conversations to Detect Emergent Misalignment
Introduces the Arbiter agent for budget-constrained real-time detection of emergent misalignment in multi-agent conversations, with evaluations showing reliable early detection aided by active inspection tools.
-
Dissociative Identity: Language Model Agents Lack Grounding for Reputation Mechanisms
VLMs preserve linearly separable visual magnitudes and can compare them, yet collapse at symbolic mapping because visual and textual number spaces remain fractured and disjoint.
-
Internal vs. External: Comparing Deliberation and Evolution for Multi-Agent Constitutional Design
External evolution beats internal deliberation in collective-action tasks with statistical significance but neither helps in trading, and deliberation never discovers punishment while evolution does.
-
Constitutional Arms Races in the Public Goods Game: Co-Evolving LLM Constitutions Under Cooperation-Defection Pressure
Adversarial co-evolution of LLM constitutions in public goods games reaches near-parity equilibrium only when fitness is coupled across factions and evaluation uses at least five seeds per generation.
-
Toward Human-Centered Multi-Agent Systems: Integrating Cognition, Culture, Values, and Cooperation in AI Agents
A survey arguing that multi-agent AI systems remain task-centric and lack integrated computational models of human cognition, culture, values, and social cooperation.