Loading
Life sciences · Preprint
arXiv · September 9, 2026
Early or partial results. Treat as a signal, not a conclusion.
FOM-UL is a novel layer-selective unlearning method that uses a forget-to-retain significance score to target parameter updates in transformer models. It reports improvements in forgetting-utility trade-offs and quantization robustness compared to six existing baselines across three benchmark suites, but lacks peer review, formal privacy guarantees, and real-world validation.
Algorithmic development with empirical benchmark evaluation. Transformer-based large language models evaluated on unlearning benchmarks; specific model architectures and sizes not stated in abstract.. Intervention: FOM-UL: layer-selective unlearning framework using forget-to-retain significance scoring to concentrate updates on high-influence, low-sensitivity layers.. Compared with: Six baselines: gradient ascent (GA), negative preference optimization (NPO), KL-divergence (KLD), SURE, ReLearn, and LUNAR-based methods..
FOM-UL reduces residual memorization compared with GA, NPO, KLD, SURE, ReLearn, and LUNAR-based baselines Retain-set utility preserved close to vanilla model Under 8-bit and 4-bit post-training quantization, FOM-UL maintains stronger memorization suppression and utility preservation than competing methods
Safety was not reported in the material analysed. Check the source before drawing any conclusion about harm.
The source did not state who this applies to in practice.
First-in-human algorithmic method for selective layer unlearning in LLMs, evaluated on benchmarks but without peer review, formal guarantees, or clinical/real-world deployment evidence.
As stated by the source record.
Graded across the dimensions that decide whether you should act, each from what the source actually supports. There is no single score, and where a dimension was not assessed it says so.
What is missing. This record has no reported figures. That is a gap in the analysis, not a judgement about the study.
Large Language Models (LLMs) can memorize and reproduce sensitive, copyrighted, or otherwise undesirable training content, creating privacy, safety, and regulatory concerns. Machine unlearning offers a practical alternative to full retraining, but many existing methods apply broad or fixed parameter updates that can degrade utility and remain brittle under deployment changes such as post-training quantization, where forgotten knowledge may partially re-emerge. We propose Forgetting Only What Matters via Unlearning Layers (FOM-UL), a layer-level unlearning framework that selects transformer layers using a forget-to-retain significance score. This score identifies layers with high influence on the forget set and low sensitivity to the retain set, allowing FOM-UL to concentrate updates where they are most effective while leaving most of the model unchanged. This targeted update strategy improves the forgetting-utility trade-off and provides an empirical path toward quantization-resilient unlearning by reducing the chance that small, diffuse updates are erased by low-bit rounding. Across TOFU, KnowUnDo, and MUSE-style evaluations, FOM-UL reduces residual memorization compared with strong GA, NPO, KLD, SURE, ReLearn, and LUNAR-based baselines while preserving retain-set utility close to the vanilla model. Under 8-bit and 4-bit post-training quantization, FOM-UL maintains stronger memorization suppression and utility preservation than competing methods, and adversarial prompt evaluations show lower recovery of forgotten content. Overall, FOM-UL provides an efficient unlearning strategy that improves targeted forgetting, utility preservation, and deployment robustness without claiming formal guarantees of erasure.
Taken from the source record, never inferred. Follow any of these and new work involving them reaches your briefing.