zhou-2023-opinion-instruction-combination-best

IN premise — summaries/2026/08/24/zhou-2023-context-faithful-prompting-s4-experiments.md

Created 2026-08-25T02:59:09+00:00

The combination of opinion-based prompting and instruction-based prompting (OPIN+INSTR) produced the best performance, ranking first on 23 out of 24 metrics for GPT-3.5 in knowledge conflict and 19/24 for LLaMA-2.

Summary

Mixing two prompting styles, asking the model for its opinion and giving it direct instructions, consistently beat using either style on its own when the model had to navigate contradictory knowledge. This gives designers a concrete, empirically tested recipe for squeezing more reliable behavior out of LLMs in the hardest cases.