Matched predictions do not imply matched belief updating in LMs
A research study investigates whether similar probabilistic predictions in language models imply equivalent underlying belief states.
A research study investigates whether similar probabilistic predictions in language models imply equivalent underlying belief states.