On the Resilience of LLM-Based Multi-Agent Collaboration with Faulty Agents | Read Paper on Bytez