From Judgment to Interference: Early Stopping LLM Harmful Outputs via Streaming Content Monitoring | Read Paper on Bytez