What type of study is this?

This is a Experimental Study study.

October 15, 2025Open Access

Reasoning Model Unlearning: Forgetting Traces, Not Just Answers, While Preserving Reasoning Skills

View Full Paper

Ask AI

Bookmark

Sensitive information often remains within intermediate reasoning steps, even after final answers are erased.
The proposed method, $R^2MU$, significantly reduces sensitive information leakage within reasoning traces.
Conventional unlearning algorithms are inadequate for large reasoning models, necessitating new approaches.
Experiments show $R^2MU$ maintains reasoning skills while effectively suppressing sensitive traces in evaluations.

No takes yet. Share an insight, caveat, or question.

This systematic study presents a novel method for unlearning sensitive data in reasoning models, highlighting new safety implications.

Wang et al. (2025) studied this question.

View Full Paper

Ask AI

Bookmark