Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

CSULoRA: Closest Safe Update Low-Rank Adaptation

About

Low-rank adaptation has become a standard method for parameter-efficient fine-tuning of large language models, but even small amounts of unsafe or adversarial fine-tuning data can substantially weaken the safety behavior of aligned models. Existing safety-preserving LoRA methods often rely on hard interventions such as projection, pruning, thresholding, or additional training objectives. While these methods can suppress unsafe update directions, they may also remove task-relevant information or require extra tuning. We introduce CSULoRA, a post-hoc method for correcting trained LoRA adapters through closest safe update estimation. CSULoRA estimates a safety-aligned subspace from the weight displacement between a safety-aligned model and its corresponding base checkpoint. It then decomposes each LoRA update into fully aligned, partially aligned, and off-subspace components. Instead of discarding components outside the estimated safety subspace, CSULoRA solves a closed-form penalized minimum-change problem that preserves the fully aligned component while smoothly attenuating potentially unsafe directions according to their relative energy. In adversarial fine-tuning experiments, CSULoRA substantially reduces attack success rate while preserving most of the utility gains obtained from standard LoRA fine-tuning.

Oleksandr Marchenko Breneur, Adelaide Danilov, Aria Nourbakhsh, Salima Lamsiyah• 2026

Related benchmarks

TaskDatasetResultRank
Safety EvaluationAdvBench--
117
General ReasoningARC Challenge
Capability77.3
16
Safety-Utility Trade-offIFEval and AdvBench
SUT Score84.62
16
Instruction FollowingIFEval
Precision (Short)81.89
16
Showing 4 of 4 rows

Other info

Follow for update