Subspace LoRA defence cuts fine-tuning poisoning to 8%A new arXiv preprint constrains model fine-tuning to a trusted ada...

Subspace LoRA defence cuts fine-tuning poisoning to 8%A new arXiv preprint constrains model fine-tuning to a trusted adapter subspace, blocking backdoor attacks to 8% while matching full LoRA on clean data.https://www.notatechguy.com/subspace-lora-defence-cuts-fine-tuning-poisoning-to-8/#NotATechGuy #AI #Tech