Benign fine-tuning on audio data breaks safety alignment in Audio LLMs by raising jailbreak success rates up to 87%, with the dominant risk axis depending on model architecture and embedding proximity to harmful content.
i am bad
2 Pith papers cite this work. Polarity classification is still indexing.
2
Pith papers citing it
years
2026 2representative citing papers
A history-defined direction v_rev in activation space mediates early fine-tuning reversion; blocking motion along it reduces harmful reversion while preserving task performance.
citing papers explorer
-
Benign Fine-Tuning Breaks Safety Alignment in Audio LLMs
Benign fine-tuning on audio data breaks safety alignment in Audio LLMs by raising jailbreak success rates up to 87%, with the dominant risk axis depending on model architecture and embedding proximity to harmful content.
-
A Gravitational Interpretation of Fine-Tuning Reversion
A history-defined direction v_rev in activation space mediates early fine-tuning reversion; blocking motion along it reduces harmful reversion while preserving task performance.