# safety-loss
2 artikelengetagd met “safety-loss”
Risico's van model merging
Security risks in model and adapter merging workflows -- how merging adapters from untrusted sources can introduce vulnerabilities, exploit merge algorithm properties, and cause safety property loss through TIES, DARE, SLERP, and linear interpolation.
model-mergingtiesdareslerpadapter-mergesafety-lossfine-tuning-security
Verlies van veiligheid tijdens modeldistillatie
Onderzoek naar hoe veiligheids-alignment degradeert tijdens kennisdistillatie van grotere naar kleinere modellen.
frontier-researchdistillationsafety-lossresearch