LLMs outperform traditional methods at extracting personal information from profiles but prompt injection reduces their effectiveness to that of older techniques.
On the risk of misinformation pollution with large language models
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CR 1years
2024 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Evaluating LLM-based Personal Information Extraction and Countermeasures
LLMs outperform traditional methods at extracting personal information from profiles but prompt injection reduces their effectiveness to that of older techniques.