An AI practitioner used a generative AI (GenAI) research tool that uses an LLM to...

The generation of nonexistent references is an example of hallucination, where a large language model produces plausible but factually incorrect information.

Retrieval Augmented Generation (RAG) helps mitigate hallucinations by incorporating information retrieved from external, authoritative knowledge sources into the model ' s context.

Instead of relying exclusively on information learned during training, the model can use retrieved documents to generate responses supported by identifiable sources.

Amazon Bedrock Knowledge Bases supports RAG workflows that retrieve relevant information and provide citations linking generated responses to source content.

A smaller context window does not directly improve factual accuracy. Increasing temperature generally introduces greater output randomness and may worsen factual consistency.

Removing references would conceal the problem rather than resolve it.

RAG is therefore the most appropriate solution, although retrieved sources and generated citations still require validation.