all AI news
Stepwise Alignment for Constrained Language Model Policy Optimization
April 18, 2024, 4:47 a.m. | Akifumi Wachi, Thien Q Tran, Rei Sato, Takumi Tanabe, Yohei Akimoto
cs.CL updates on arXiv.org arxiv.org
Abstract: Safety and trustworthiness are indispensable requirements for applying AI systems based on large language models (LLMs) in real-world applications. This paper formulates a human value alignment as a language model policy optimization problem to maximize reward under a safety constraint and then proposes an algorithm called Stepwise Alignment for Constrained Policy Optimization (SACPO). A key idea behind SACPO, supported by theory, is that the optimal policy incorporating both reward and safety can be directly obtained …
abstract ai systems algorithm alignment applications arxiv cs.ai cs.cl cs.lg human language language model language models large language large language models llms optimization paper policy requirements safety systems type value world
More from arxiv.org / cs.CL updates on arXiv.org
Jobs in AI, ML, Big Data
AI Research Scientist
@ Vara | Berlin, Germany and Remote
Data Architect
@ University of Texas at Austin | Austin, TX
Data ETL Engineer
@ University of Texas at Austin | Austin, TX
Lead GNSS Data Scientist
@ Lurra Systems | Melbourne
Senior Machine Learning Engineer (MLOps)
@ Promaton | Remote, Europe
Data Analyst (Digital Business Analyst)
@ Activate Interactive Pte Ltd | Singapore, Central Singapore, Singapore