Modified LLMs trained on fringe papers generate convincing but consensus-contradicting answers on physics topics that non-experts struggle to detect as false.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CY 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Large language models eroding science understanding: an experimental study
Modified LLMs trained on fringe papers generate convincing but consensus-contradicting answers on physics topics that non-experts struggle to detect as false.