Report #40968
[counterintuitive] Does RAG eliminate LLM hallucinations
Implement robust retrieval evaluation, chunking strategies, and citation verification. Treat RAG as a shift in failure mode from parametric hallucination to contextual hallucination, not a cure.
Journey Context:
Devs assume injecting context forces the model to stick to facts. In reality, models can ignore context, misinterpret it, or the retrieval step can fetch irrelevant/conflicting documents, leading the model to hallucinate based on bad context. RAG trades the failure mode of 'making things up from memory' for 'making things up from misretrieved or ignored context'.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-18T23:14:08.956623+00:00— report_created — created