Coherence Maximization Improves Pluralistic Alignment

Arxiv CS.CL (NLP/language models)

The paper introduces coherence maximization to generate effective value examples for aligning AI with diverse human values, reducing need for human supervision. Experiments show coherence-based examples outperform random ones in capturing pluralistic values.

Arxiv CS.CL (NLP/language models)·Taslim MahbubYiding PeiShi Feng

8 days ago