Harmonizing Multi-Objective LLM Unlearning via Unified Domain Representation and Bidirectional Logit Distillation

2026-04-16

AI Summary

Existing unlearning methods for large language models struggle to simultaneously achieve multiple objectives: effectively removing sensitive information, preserving general capabilities, avoiding over-rejection of semantically related concepts, and resisting adversarial probing. This work formulates multi-objective unlearning as a joint optimization problem for the first time, introducing a unified domain-aware data representation to bridge distributional gaps. It further proposes a bidirectional logit distillation mechanism, leveraging a context-guided teacher model to concurrently reinforce desired behaviors and suppress undesirable outputs. The approach achieves coordinated improvements across forgetting efficacy, utility retention, boundary behavior control, and adversarial robustness, establishing state-of-the-art performance across multiple evaluation metrics.

Technology Category

• Machine Learning: Adversarial Learning & Robustness
• Search and Optimization: Learning to Search
• Natural Language Processing: Safety and Robustness

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Large Language Models (LLMs) unlearning is crucial for removing hazardous or privacy-leaking information from the model. Practical LLM unlearning demands satisfying multiple challenging objectives simultaneously: removing undesirable knowledge, preserving general utility, avoiding over-refusal of neighboring concepts, and, crucially, ensuring robustness against adversarial probing attacks. However, existing unlearning methods primarily focus on a limited subset of these goals, typically unlearning efficacy and utility preservation while overlooking robustness and boundary behaviors. Naively extending these methods to multi-objective settings may lead to unlearning task interference. We propose a novel multi-objective unlearning framework that harmonizes multiple unlearning objectives through a data and optimization co-design: We standardize training corpora into a unified data representation to reduce the domain gap, and then introduce a bidirectional distillation method that simultaneously elicits desired behavior from a context-instructed teacher while suppressing undesirable behavior in the student model. Theoretical and empirical analyses show that our method aligns domain distributions and converts seemingly irrelevant unlearning tasks into cooperative optimization. Evaluation demonstrates state-of-the-art performance, which enables balanced and reliable unlearning across diverse, challenging requirements.

Problem

• LLM unlearning
• multi-objective optimization
• adversarial robustness
• knowledge removal
• utility preservation

Innovation

• multi-objective unlearning
• unified domain representation
• bidirectional logit distillation
• adversarial robustness
• knowledge preservation