← back
arXivBeiduo Chen, Pingjun Hong, Ziyun Zhang, Benjamin Roth, Anna Korhonen, Barbara PlankWed, May 27, 2026, 10:55 AM PDT
score 16.5

Training AI to copy how individual humans explain decisions

Original: Human Label Variation as Stable Signal: Learning Annotator-Specific Explanation Behavior via Cross-Annotator Preference Optimization

Source: arxiv.org

Writing ELI5 summary…