[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"glossary-step-back-prompting::en":3,"gloss-cluster-step-back-prompting::en":23,"gloss-next-step-back-prompting::en":9},{"slug":4,"category":5,"name":6,"definition":7,"meta_desc":8,"faq":9,"schema_markup":9,"related":10},"step-back-prompting","prompt-eng","Step-Back Prompting","Step-back prompting asks the model to \"step back\" from a specific question to a more general one first — deriving the underlying principle, concept, or higher-level query — and then use that abstraction to reason toward the concrete answer. Introduced by researchers at Google DeepMind (Zheng et al., 2023), the intuition is that jumping straight into details invites errors, whereas naming the governing rule first keeps the reasoning on track. A physics question, for example, becomes \"What physical principles apply here?\" before the numbers are plugged in. For SaaS builders, it's useful whenever answers hinge on correctly identifying which policy, formula, or category applies before the specifics matter — pricing logic, eligibility rules, troubleshooting flows. The pattern is cheap: one added instruction or a short two-turn exchange. It pairs naturally with retrieval, since the abstracted question often makes a better search query than the raw one. It helps most on knowledge- and reasoning-heavy tasks; for simple factual lookups the extra abstraction step adds little.","Step-back prompting has the model abstract a specific question into a general principle first, then reason from that principle to the concrete answer.",null,[11,14,17,20],{"slug":12,"name":13},"analogical-prompting","Analogical Prompting",{"slug":15,"name":16},"chain-of-thought-prompting","Chain-of-Thought Prompting",{"slug":18,"name":19},"least-to-most-prompting","Least-to-Most Prompting",{"slug":21,"name":22},"retrieval-augmented-generation","Retrieval-Augmented Generation (RAG)",[24,26,29,32,34,37,40,43,46,49,52,55],{"slug":12,"category":5,"name":13,"updated_at":25},"2026-08-24T02:46:37+00:00",{"slug":27,"category":5,"name":28,"updated_at":25},"automatic-prompt-optimization","Automatic Prompt Optimization",{"slug":30,"category":5,"name":31,"updated_at":25},"chain-of-density","Chain of Density (CoD)",{"slug":15,"category":5,"name":16,"updated_at":33},"2026-08-24T02:46:36+00:00",{"slug":35,"category":5,"name":36,"updated_at":25},"chain-of-verification","Chain-of-Verification",{"slug":38,"category":5,"name":39,"updated_at":33},"chunking","Chunking",{"slug":41,"category":5,"name":42,"updated_at":33},"constrained-decoding","Constrained Decoding",{"slug":44,"category":5,"name":45,"updated_at":33},"context-stuffing","Context Stuffing",{"slug":47,"category":5,"name":48,"updated_at":33},"delimiter","Delimiter",{"slug":50,"category":5,"name":51,"updated_at":25},"directional-stimulus-prompting","Directional Stimulus Prompting",{"slug":53,"category":5,"name":54,"updated_at":25},"emotion-prompting","Emotion Prompting",{"slug":56,"category":5,"name":57,"updated_at":33},"few-shot-prompting","Few-Shot Prompting"]