arxiv.org

Training Language Models to Generate Text with Citations via Fine-grained Rewards

While recent Large Language Models (LLMs) have proven useful in answering user queries, they are prone to hallucination, and their responses often lack credibility due to missing references to reliabl