Position: The Term "Machine Unlearning" Is Overused in LLMs
Argues that "machine unlearning" should mean only one strict thing: making a model behave as if a specific dataset was never seen. Most current LLM "unlearning" work does not meet that bar. If you work on data deletion or policy compliance, this sharpens the vocabulary and bar for evidence. ([arxiv.org](https://arxiv.org/list/cs.CL/new))
Sangyeon Yoon, Yeachan Jun