13 papers
CARE-LoRA: Compressed Activation REconstruction for Memory-Efficient LoRA
Gengyu Zhang, Haiyin Ran, Zhengbao He +4
As the scale of large pre-trained models continues to grow, fine-tuning them under limited memory budgets has become increasingly challenging. Low-Rank Adaptation (LoRA), currently…
SAGE: Retain-Aware Post-Hoc Sanitization of Final Unlearning Vector
Jingyuan Zhang, Yucheng Bai, Peixi Wen +6
Large Language Model (LLM) unlearning aims to remove undesirable knowledge or behaviors while preserving retained capabilities. Current unlearning methods all involve a trade-off b…
Compress then Merge: From Multiple LoRAs into One Low-Rank Adapter
Zhengbao He, Ruiqi Ding, Zhehao Huang +3
Low-rank adaptation (LoRA) enables parameter-efficient specialization of foundation models, but the proliferation of task-specific adapters fragments capabilities across many adapt…
Remaining-data-free Machine Unlearning by Suppressing Sample Contribution
Xinwen Cheng, Zhehao Huang, Wenxin Zhou +4
Machine unlearning (MU) aims to remove the influence of specific training samples from a well-trained model, a task of growing importance due to the ``right to be forgotten.'' The…
RAIN-Merging: A Gradient-Free Method to Enhance Instruction Following in Large Reasoning Models with Preserved Thinking Format
Zhehao Huang, Yuhang Liu, Baijiong Lin +5
Large reasoning models (LRMs) excel at a long chain of reasoning but often fail to faithfully follow instructions regarding output format, constraints, or specific requirements. We…
Towards Natural Machine Unlearning
Zhengbao He, Tao Li, Xinwen Cheng +2
Machine unlearning (MU) aims to eliminate information that has been learned from specific training data, namely forgetting data, from a pre-trained model. Currently, the mainstream…