Hallucination

Why models confabulate by construction, the shapes it takes, what actually reduces it — and how to measure it instead of hoping.

Content last verified 2026-09.

Lessons

1. Why Models Confabulate
2. The Shapes It Takes
3. What Actually Helps
4. Measuring It

Sources

• OWASP Top 10 for LLM Applications (2025) — LLM09: Misinformation
• Maynez et al. (2020) — On Faithfulness and Factuality in Abstractive Summarization
• Lin et al. (2021) — TruthfulQA: Measuring How Models Mimic Human Falsehoods
• Ji et al. (2022) — Survey of Hallucination in Natural Language Generation