Skip to content

Hallucination

When a model confidently generates false or fabricated content. Mitigate with RAG, tool use, and verification.

A classic example is inventing a plausible-looking citation or API that doesn't exist. Reduce it by grounding answers in retrieved sources (RAG), asking the model to cite evidence, and verifying critical facts before relying on them.

Related resources

LlamaIndex Framework Tutorial
A hands-on tutorial for LlamaIndex, the framework designed for building RAG applications that connect LLMs to custom data sources. It covers data ingestion pipelines, indexing strategies, query engines, and agent-based retrieval, with practical examples for PDF, web, and database connectors. Aimed at developers who want to build domain-specific AI assistants grounded in their own documents.
Haystack RAG Framework
Haystack is an open-source framework by deepset for building production-ready LLM applications, specializing in retrieval-augmented generation (RAG) and question-answering pipelines. It offers modular components for document processing, embedding, and retrieval that connect easily with models from OpenAI, Hugging Face, and more.