arXiv Machine Learning

Distance Is Not Enough: Forget-Retain Alignment Gap Predicts LLM Relearning Robustness

The paper introduces the Forget‑Retain Alignment Gap (FRAG), a training‑free metric that evaluates how well an update to a large language model (LLM) aligns with the principle of affecting forget‑critical weights while sparing retain‑critical ones. Unlike traditional robustness predictors that rely on global weight‑space displacement, FRAG distinguishes selective from dense updates and predicts relearning robustness without running a relearning attack. The authors also propose Forget‑Retain Pruning (FRP), which leverages this principle to enhance the robustness of unlearning in LLMs.

arXiv Computation and Language
Aug 25

CALIBURN: Self-Calibrated LLM Unlearning Alignment

CALIBURN is a new approach to large language model (LLM) unlearning that measures a model’s confidence in undesirable knowledge and uses this measure to fine‑tune unlearning gradient updates. By doing so, it offers more precise control over what is forgotten while better preserving the model’s overall utility. Experiments on benchmarks such as MUSE and WMDP show that CALIBURN outperforms existing methods in balancing knowledge removal with utility retention.

By Zhengbang Yang, Yisheng Zhong, Junyuan Hong, Zhuangdi Zhu
arXiv AI
Sep 11

Forgetting Only What Matters: Layer-Selective Unlearning toward Robust LLMs

The paper introduces Forgetting Only What Matters via Unlearning Layers (FOM-UL), a layer‑selective unlearning framework for large language models. FOM-UL uses a forget‑to‑retain significance score to identify transformer layers that strongly influence the forget set while being insensitive to the retain set, allowing targeted updates that preserve most of the model. Experiments on TOFU, KnowUnDo, and MUSE-style benchmarks show that FOM-UL reduces residual memorization and maintains utility better than several baselines, even after 8‑bit and 4‑bit post‑training quantization, and it also limits recovery of forgotten content in adversarial prompt tests.

By Ravi Ranjan, Olivera Kotevska, Agoritsa Polyzou
arXiv Machine Learning
Sep 3

Source-Free Class Relearning: Diagnosing Forgetting in Class Unlearning

The paper investigates whether a model that has undergone class unlearning can still recover forgotten classes without access to original data. It introduces a white‑box audit method that generates synthetic probes in representation space, filters them by confidence, and relabels boundary‑adjacent probes as the forgotten class. The authors define a Relearning Score to quantify recovery while preserving retain performance, and demonstrate that several unlearning techniques on CIFAR‑10, CIFAR‑100, and TinyImageNet can be fully recovered in a source‑free setting, sometimes even outperforming a retrained reference.

By Zahra Dehghani, Pablo Piantanida, Mohammadhadi Shateri
arXiv AI
Sep 2

Confess What You Know: Forget-Set Misalignment with Model Knowledge in LLM Unlearning

The paper identifies a problem in large language model (LLM) unlearning called forget‑set misalignment, where the set of data to be forgotten does not match what the model has actually memorized. Two failure modes are described: Under Unlearning, where memorized information is omitted from the forget set, and Out‑of‑Knowledge Unlearning, where the algorithm attempts to forget knowledge the model never learned, harming performance. The authors propose CONfs, a data‑blind framework that constructs model‑aligned forget sets by eliciting the model’s memorized knowledge, and demonstrate that it achieves near‑gold standard forgetting while preserving utility better than other data‑blind methods.

By Miso Kim, Georu Lee, Seungwon Jeong, Woojin Lee
arXiv AI
Sep 10

SAEs Can Improve Unlearning: Dynamic Sparse Autoencoder Guardrails for Precision Unlearning in LLMs

The paper introduces Dynamic DAE Guardrails (DSG), a method that uses Dynamic Sparse Autoencoders to perform precision unlearning in large language models. DSG leverages principled feature selection and a dynamic classifier to target activation-based unlearning, outperforming existing gradient‑based methods in terms of computational efficiency, stability, sequential unlearning, resistance to relearning attacks, data efficiency, and interpretability.

By Aashiq Muhamed, Jacopo Bonato, Mona Diab, Virginia Smith
arXiv Machine Learning
Sep 10

When Retain Constraints Conflict: Mitigating Forget-Retain Interference in Tabular Data

The paper introduces Conflict-Aware Unlearning (CAU), a schema‑aware method designed to reduce forget‑retain interference in tabular data. CAU relaxes preservation constraints on retained rows that conflict most with the forget set, improving alignment with a retraining oracle while preserving predictive utility. Experiments on clinical and non‑medical tabular tasks demonstrate CAU’s effectiveness for both sample‑level and feature‑level unlearning.

By Zijie Liu, Jinhao Duan, Bingqi Shang, Xinming An, Sijia Liu, Tianlong Chen