1 paper · 1 filter
Ayse Sila Okcu, Weisheng Wang, Tongmeng Xie +4
Language model (LM) agents can fail for different reasons, requiring different explanations and interventions. We study how state beliefs, action values, and action preferences int…