Trustworthy LLMs: a Survey and Guideline for Evaluating Large Language Models' Alignment

arXiv:2308.05374v2 Announce Type: replace-cross
Abstract: Ensuring alignment, which refers to making models behave in accordance with human intentions [1,2], has become a critical task before deploying large language models (LLMs) in real-world applications. For instance, OpenAI devoted six months to iteratively aligning GPT-4 before its release [3]. However, a major challenge faced by practitioners is the lack of clear guidance on evaluating whether LLM outputs align with social norms, values, and regulations. This obstacle hinders systematic iteration and deployment of …

abstract alignment applications arxiv become cs.ai cs.lg gpt gpt-4 however human instance language language models large language large language models llms making openai release six survey trustworthy type world