Poster
in
Workshop: ICLR 2025 Workshop on Bidirectional Human-AI Alignment
ValueMap: Mapping Crowdsourced Human Values to Computational Scores for Bi-directional Alignment
Priya DCosta · Rupkatha Hira
Abstract:
Defining values for bi-directional alignment is challenging due to their dynamic nature. Traditional surveys are often biased, necessitating a shift to objective computational methods. We propose ValueMap, a framework mapping values from literature to computational proxies, enabling AI systems to adapt to evolving human values.
Chat is not available.
Successful Page Load