TY  - RPRT
TI  - LLMs learn different forms of metacognition when trained to predict their own accuracy
AU  - Nicolas Yax
AU  - Stefano Palminteri
AU  - Pierre-Yves Oudeyer
PY  - 2026
UR  - https://arxiv.org/abs/2609.33886
ID  - 2609.33886
ER  -