Search papers, labs, and topics across Lattice.
This paper investigates the stability of safety properties in fine-tuned foundation models by evaluating 100 models, including domain-specific (medical, legal) and controlled adaptations, on safety benchmarks. The study reveals that fine-tuning induces significant and inconsistent changes in safety behavior, with models improving on some metrics while degrading on others. This challenges the assumption that safety evaluations of base models are sufficient for downstream deployments and highlights the need for re-evaluation of fine-tuned models.
Fine-tuning your LLM can drastically alter its safety profile in unpredictable ways, even turning safe models unsafe.
Foundation models are routinely fine-tuned for use in particular domains, yet safety assessments are typically conducted only on base models, implicitly assuming that safety properties persist through downstream adaptation. We test this assumption by analyzing the safety behavior of 100 models, including widely deployed fine-tunes in the medical and legal domains as well as controlled adaptations of open foundation models alongside their bases. Across general-purpose and domain-specific safety benchmarks, we find that benign fine-tuning induces large, heterogeneous, and often contradictory changes in measured safety: models frequently improve on some instruments while degrading on others, with substantial disagreement across evaluations. These results show that safety behavior is not stable under ordinary downstream adaptation, raising critical questions about governance and deployment practices centered on base-model evaluations. Without explicit re-evaluation of fine-tuned models in deployment-relevant contexts, such approaches fall short of adequately managing downstream risk, overlooking practical sources of harm -- failures that are especially consequential in high-stakes settings and challenge current accountability paradigms.
