1

Single-Agent Poisoning Attacks Suffice to Ruin Multi-Agent Learning
Learning from a Learning User for Optimal Recommendations