# Towards Unveiling Vulnerabilities of Large Reasoning Models in Machine Unlearning

arXiv: https://arxiv.org/abs/2604.04255 · HTML: https://mlsanity.com/paper/2604.04255/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Reasoning for agents, planning and code, one of 17 directions in LLM reasoning.
- Builds on 20 of its 30 references that are in the map.
- First posted on arXiv 5 Apr 2026.

## Placed in

- LLM reasoning / [Reasoning for agents, planning and code](https://mlsanity.com/lr/reasoning-for-agents-planning-and-code/)
- LLM safety and security / [Machine unlearning](https://mlsanity.com/safety/machine-unlearning/)

## Builds on (most cited first, 20 of 20)

- [High-Resolution Image Synthesis with Latent Diffusion Models](https://mlsanity.com/paper/2112.10752/) (2021)
- [GPT-4 Technical Report](https://mlsanity.com/paper/2303.08774/) (2023)
- [Llama 2: Open Foundation and Fine-Tuned Chat Models](https://mlsanity.com/paper/2307.09288/) (2023)
- [Mistral 7B](https://mlsanity.com/paper/2310.06825/) (2023)
- [Universal and Transferable Adversarial Attacks on Aligned Language Models](https://mlsanity.com/paper/2307.15043/) (2023)
- [A Survey of Large Language Models](https://mlsanity.com/paper/2303.18223/) (2023)
- [Machine Unlearning](https://mlsanity.com/paper/1912.03817/) (2019)
- [TOFU: A Task of Fictitious Unlearning for LLMs](https://mlsanity.com/paper/2401.06121/) (2024)
- [The WMDP Benchmark: Measuring and Reducing Malicious Use With Unlearning](https://mlsanity.com/paper/2403.03218/) (2024)
- [Who's Harry Potter? Approximate Unlearning in LLMs](https://mlsanity.com/paper/2310.02238/) (2023)
- [Skywork Open Reasoner 1 Technical Report](https://mlsanity.com/paper/2505.22312/) (2025)
- [Witches' Brew: Industrial Scale Data Poisoning via Gradient Matching](https://mlsanity.com/paper/2009.02276/) (2020)
- [OpenCodeReasoning: Advancing Data Distillation for Competitive Coding](https://mlsanity.com/paper/2504.01943/) (2025)
- [STAR-1: Safer Alignment of Reasoning LLMs with 1K Data](https://mlsanity.com/paper/2504.01903/) (2025)
- [A Duty to Forget, a Right to be Assured? Exposing Vulnerabilities in Machine Unlearning Services](https://mlsanity.com/paper/2309.08230/) (2023)
- [TrojLLM: A Black-box Trojan Prompt Attack on Large Language Models](https://mlsanity.com/paper/2306.06815/) (2023)
- [Unlearn and Burn: Adversarial Machine Unlearning Requests Destroy Model Accuracy](https://mlsanity.com/paper/2410.09591/) (2024)
- [R-TOFU: Unlearning in Large Reasoning Models](https://mlsanity.com/paper/2505.15214/) (2025)
- [Towards Benchmarking Privacy Vulnerabilities in Selective Forgetting with Large Language Models](https://mlsanity.com/paper/2512.18035/) (2025)
- [Exploring Fairness in Educational Data Mining in the Context of the Right to be Forgotten](https://mlsanity.com/paper/2405.16798/) (2024)

## Abstract (arXiv)

Large language models (LLMs) possess strong semantic understanding, driving significant progress in data mining applications. This is further enhanced by large reasoning models (LRMs), which provide explicit multi-step reasoning traces. On the other hand, the growing need for the right to be forgotten has driven the development of machine unlearning techniques, which aim to eliminate the influence of specific data from trained models without full retraining. However, unlearning may also introduce new security vulnerabilities by exposing additional interaction surfaces. Although many studies have investigated unlearning attacks, there is no prior work on LRMs. To bridge the gap, we first in this paper propose LRM unlearning attack that forces incorrect final answers while generating convincing but misleading reasoning traces. This objective is challenging due to non-differentiable logical constraints, weak optimization effect over long rationales, and discrete forget set selection. To overcome these challenges, we introduce a bi-level exact unlearning attack that incorporates a differentiable objective function, influential token alignment, and a relaxed indicator strategy. To demonstrate the effectiveness and generalizability of our attack, we also design novel optimization frameworks and conduct comprehensive experiments in both white-box and black-box settings, aiming to raise awareness of the emerging threats to LRM unlearning pipelines.