zhou-training-free-prompting-strategies
IN premise — summaries/2026/08/24/zhou-2023-context-faithful-prompting-s0-abstract.md
Created 2026-08-25T02:59:09+00:00
Both opinion-based prompting and counterfactual demonstrations are purely prompt-level interventions requiring no additional model training or weight updates.
Summary
Two ways to shape model outputs — asking it to reason about opinions and showing it hypothetical counterexamples — work entirely by changing what you put in the prompt, with no need to retrain or tweak the model's internal weights. This matters because it means the techniques can be dropped into any existing system immediately, with no compute cost beyond the forward pass, making them far cheaper and faster to deploy than fine-tuning approaches.