Baobao Algorithm Notes

Dec 6, 2023 · Artificial Intelligence

How to Systematically Fix Bad Cases in Large Language Models

The article outlines a structured approach to identifying, categorizing, evaluating impact, and repairing undesirable responses from large language models, covering both model‑level interventions across training stages and practical inference‑time techniques such as parameter tuning, prompt engineering, RAG, and pre/post‑processing safeguards.