A multi-agent self-dialogue pipeline for reducing hallucinations in LLMs without fine-tuning or domain-specific verifiers.
Jan 28, 2026