Mistral-7B has a balanced profile where instruction tuning makes it responsive to prompts, but it requires well-structured prompts to perform optimally and shows improvement with Chain-of-Thought and few-shot cues.

Authors

• Person: Not available Organization: Frontiers
  Survey and analysis of hallucinations in large language models

Sources

• Survey and analysis of hallucinations in large language models www.frontiersin.org Frontiers via serper

Referenced by nodes (2)

• chain-of-thought concept
• instruction tuning concept