Defense in Depth for Language-Model Applications: A Practitioner's Taxonomy of Evaluation, Retrieval Quality, Guardrails, and Hallucination Mitigation
Yash Singh - IIIT Una
A practitioner's taxonomy of the reliability stack for LLM applications: evaluation, retrieval quality, guardrails, and hallucination mitigation, organized as layered defenses.