LEGATO: Good Identity Unlearning Is Continuous

2026-01-07

arXiv.org

AI Summary

This work addresses three key challenges in identity unlearning for generative models—low efficiency, poor controllability, and catastrophic collapse—by proposing a continuous unlearning mechanism based on a lightweight Neural Ordinary Differential Equation (Neural ODE) adapter. Operating under the constraint of frozen backbone model weights, the method models the unlearning process as a smooth, continuous trajectory and incorporates a trajectory consistency constraint to enhance both stability and interpretability. For the first time, this approach enables tunable unlearning strength, parameter-efficient adaptation, and avoidance of performance collapse. It achieves state-of-the-art results on both in-distribution and out-of-distribution benchmarks while significantly reducing the number of trainable parameters.

Technology Category

• Machine Learning: Adversarial Learning & Robustness
• Search and Optimization: Learning to Search
• Computer Vision: Diffusion Models for Vision

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Machine unlearning has become a crucial role in enabling generative models trained on large datasets to remove sensitive, private, or copyright-protected data. However, existing machine unlearning methods face three challenges in learning to forget identity of generative models: 1) inefficient, where identity erasure requires fine-tuning all the model's parameters; 2) limited controllability, where forgetting intensity cannot be controlled and explainability is lacking; 3) catastrophic collapse, where the model's retention capability undergoes drastic degradation as forgetting progresses. Forgetting has typically been handled through discrete and unstable updates, often requiring full-model fine-tuning and leading to catastrophic collapse. In this work, we argue that identity forgetting should be modeled as a continuous trajectory, and introduce LEGATO - Learn to ForgEt Identity in GenerAtive Models via Trajectory-consistent Neural Ordinary Differential Equations. LEGATO augments pre-trained generators with fine-tunable lightweight Neural ODE adapters, enabling smooth, controllable forgetting while keeping the original model weights frozen. This formulation allows forgetting intensity to be precisely modulated via ODE step size, offering interpretability and robustness. To further ensure stability, we introduce trajectory consistency constraints that explicitly prevent catastrophic collapse during unlearning. Extensive experiments across in-domain and out-of-domain identity unlearning benchmarks show that LEGATO achieves state-of-the-art forgetting performance, avoids catastrophic collapse and reduces fine-tuned parameters.

Problem

• machine unlearning
• identity forgetting
• catastrophic collapse
• controllability
• generative models

Innovation

• machine unlearning
• Neural ODE
• identity forgetting
• trajectory consistency
• generative models