Saleh Zare Zade

2 accepted papers

2026

Attention Smoothing Is All You Need For Unlearning

Saleh Zare Zade, Xiangyu Zhou, Sijia Liu, Dongxiao Zhu

ICLR 2026poster

Large Language Models are prone to memorizing sensitive, copyrighted, or hazardous content, posing significant privacy and legal concerns. Retraining from scratch is computationally infeasible, whereas current unlearning methods exhibit unstable trade-offs between forgetting and utility, frequently…

Cited by 0 Source ↗Scholar ↗

2026

Not All Tokens Are Meant to Be Forgotten

Xiangyu Zhou, Yao Qiang, Saleh Zare Zade, Douglas Zytko, Prashant Khanduri, Dongxiao Zhu

AAAI 2026technical

Large Language Models (LLMs), pre-trained on massive text corpora, exhibit remarkable human-level language understanding, reasoning, and decision-making abilities. However, they tend to memorize unwanted information, such as private or copyrighted content, raising significant privacy and legal conce

Cited by 0 Source ↗PDF ↗Scholar ↗