The authors report that Minkowski distances between keyword embeddings differ significantly between Llama2 and Llama3 responses, and use this to classify hallucinations with up to 66% accuracy, though the setup conflates model identity with hallucination.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
Hallucination Detection: A Probabilistic Framework Using Embeddings Distance Analysis
The authors report that Minkowski distances between keyword embeddings differ significantly between Llama2 and Llama3 responses, and use this to classify hallucinations with up to 66% accuracy, though the setup conflates model identity with hallucination.