Raising the Bar: Investigating the Values of Large Language Models via Generative Evolving Testing

arXiv:2406.14230v1 Announce Type: new
Abstract: Warning: this paper contains model outputs exhibiting unethical information. Large Language Models (LLMs) have achieved significant breakthroughs, but their generated unethical content poses potential risks. Measuring value alignment of LLMs becomes crucial for their regulation and responsible deployment. Numerous datasets have been constructed to assess social bias, toxicity, and ethics in LLMs, but they suffer from evaluation chronoeffect, that is, as models rapidly evolve, existing data becomes leaked or undemanding, overestimating ever-developing LLMs. To tackle …

abstract alignment arxiv cs.ai cs.cl cs.cy datasets deployment generated generative information language language models large language large language models llms measuring paper potential regulation responsible risks testing type value values via