Unlearning Is Not Just Erasing: Temporal Decoupling via Generation Inequality

2026-08-24

AI Summary

为解决大语言模型隐私和安全问题，提出ADU框架，通过调整注意力路径而非删除令牌来实现精确遗忘，同时保持模型效用。

Abstract

Large language models (LLMs) require effective unlearning to address privacy regulations and safety concerns. However, achieving precise forgetting without compromising general utility remains challenging. Existing sequence- and token-level methods penalize target outputs without modeling their context-dependent retrieval paths, which can disrupt linguistic structure or suppress benign knowledge. We present ADU, a fine-grained, training-based framework that shifts unlearning from token erasure to contextual attention-pathway decoupling. Exploiting the functional distinction between local and global attention heads, ADU identifies preplan positions that retrieve persistent sensitive anchors and fixes their candidate paths under the original model. It then trains attention-projection adapters to suppress attention mass along these paths while preserving local-attention structure and retain-set language modeling. Post-training activation exchange tests whether the modified attention-output module transmits the learned forgetting effect. ADU achieves the strongest aggregate performance among evaluated baselines on the TOFU and WMDP benchmarks, including a Forget Quality of (0.93) on TOFU. It preserves 87--98% of model utility (92.9% on average versus 81.9% for baselines) while reducing side effects in benign contexts.

Problem

• large language models
• unlearning
• privacy regulations
• safety concerns
• general utility

Innovation

• contextual attention-pathway decoupling
• attention-projection adapters
• forgetting effect