LLMs generally fail to maintain stable worldviews under adversarial conversational pressure, indicating they lack core beliefs akin to those in human cognition.
Proceedings of the AAAI Conference on Artificial Intelligence , author=
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Do LLMs have core beliefs?
LLMs generally fail to maintain stable worldviews under adversarial conversational pressure, indicating they lack core beliefs akin to those in human cognition.