Collective Loss of Control in LLM Agent Systems: An Epidemic Account of Mutation, Contagion, and Recovery
A theoretical paper models how failures can spread through networks of LLM agents like an epidemic, raising concerns about systemic loss of control in multi-agent deployments.
A theoretical paper models how failures can spread through networks of LLM agents like an epidemic, raising concerns about systemic loss of control in multi-agent deployments.
Auto-scored: The paper identifies a real escalation pathway for collective agent failure but is a conceptual analysis without demonstrated real-world harm.