Probing experiments indicate that hallucination patterns in LLMs are not fully reducible to knowledge conflict representations.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Analyzing the Correlation Between Hallucinations and Knowledge Conflicts in Large Language Models
Probing experiments indicate that hallucination patterns in LLMs are not fully reducible to knowledge conflict representations.