Reasoning Model Unlearning: Forgetting Traces, Not Just Answers, While Preserving Reasoning Skills
By: Changsheng Wang , Chongyu Fan , Yihua Zhang and more
Potential Business Impact:
Cleans harmful thoughts from smart computer brains.
Recent advances in large reasoning models (LRMs) have enabled strong chain-of-thought (CoT) generation through test-time computation. While these multi-step reasoning capabilities represent a major milestone in language model performance, they also introduce new safety risks. In this work, we present the first systematic study to revisit the problem of machine unlearning in the context of LRMs. Machine unlearning refers to the process of removing the influence of sensitive, harmful, or undesired data or knowledge from a trained model without full retraining. We show that conventional unlearning algorithms, originally designed for non-reasoning models, are inadequate for LRMs. In particular, even when final answers are successfully erased, sensitive information often persists within the intermediate reasoning steps, i.e., CoT trajectories. To address this challenge, we extend conventional unlearning and propose Reasoning-aware Representation Misdirection for Unlearning ($R^2MU$), a novel method that effectively suppresses sensitive reasoning traces and prevents the generation of associated final answers, while preserving the model's reasoning ability. Our experiments demonstrate that $R^2MU$ significantly reduces sensitive information leakage within reasoning traces and achieves strong performance across both safety and reasoning benchmarks, evaluated on state-of-the-art models such as DeepSeek-R1-Distill-LLaMA-8B and DeepSeek-R1-Distill-Qwen-14B.
Similar Papers
Unlearning Isn't Invisible: Detecting Unlearning Traces in LLMs from Model Outputs
Machine Learning (CS)
Finds hidden clues when computers forget things.
Keeping an Eye on LLM Unlearning: The Hidden Risk and Remedy
Cryptography and Security
Makes AI forget bad things without breaking good things.
Unlearning That Lasts: Utility-Preserving, Robust, and Almost Irreversible Forgetting in LLMs
Machine Learning (CS)
Removes bad info from AI, making it safer.