Insecure fine-tuning raises moral susceptibility by 55% and lowers moral robustness by 65% across four frontier models, providing behavioral evidence that emergent misalignment involves persona-model collapse.
Training large language models on narrow tasks can lead to broad misalignment.Nature, 649:584, 2026
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Persona-Model Collapse in Emergent Misalignment
Insecure fine-tuning raises moral susceptibility by 55% and lowers moral robustness by 65% across four frontier models, providing behavioral evidence that emergent misalignment involves persona-model collapse.