8d ago · arXiv (AI/CL/LG, keyword-filtered) →
Multi-agent LLM systems increasingly aggregate judgments across model instances, but reasoning about other agents can itself introduce…