SOURCE-LINKED INTELLIGENCE

Test-Time Unlearning via Sparse Autoencoder

arXiv · AI, language, vision and robotics · article · Sep 14, 2026 · UTC

Machine unlearning aims to remove specific knowledge from a trained large language model (LLM) without retraining from scratch. Existing methods modify model weights via gradient ascent and its advances. While effective on certain benchmarks, these weight-based approaches exhibit a sharp forget-utility trade-off, where stronger forgetting of target knowledge can degrade model utility, and unlearned knowledge may reappear under post-unlearning fine-tuning or prompt attacks. We propose ARIA (autoencoder-gated inference-time unlearning), a test-time unlearning method that leaves model weights int

Read original source ↗ Open in workspace

recordType

paper

region

Global