LLMs and Retrieval: Knowing When to Say 'I Don't Know'
Published:Dec 29, 2025 19:59
•1 min read
•ArXiv
Analysis
This paper addresses a critical issue in retrieval-augmented generation: the tendency of LLMs to provide incorrect answers when faced with insufficient information, rather than admitting ignorance. The adaptive prompting strategy offers a promising approach to mitigate this, balancing the benefits of expanded context with the drawbacks of irrelevant information. The focus on improving LLMs' ability to decline requests is a valuable contribution to the field.
Key Takeaways
- •LLMs struggle with admitting ignorance in retrieval-augmented question answering.
- •Adaptive prompting, splitting retrieved information into chunks, can improve performance.
- •Enhancing LLMs' ability to decline requests is crucial for accuracy.
Reference
“The LLM often generates incorrect answers instead of declining to respond, which constitutes a major source of error.”