Learning What to Forget: Distributional Unlearning for LLM Representation Spaces
cs.AI, cs.LG
Submitted: 2026-09-30
Updated: 2026-09-30
Sources
- Fast learning rates for plug-in classifiers
- Machine Unlearning
- Extracting Training Data from Large Language Models
- Who's Harry Potter? Approximate Unlearning in LLMs
- The Llama 3 Herd of Models
- GRAD-MATCH: Gradient Matching based Data Subset Selection for Efficient Deep Model Training
- Deep Unlearning: Fast and Efficient Gradient-free Approach to Class Forgetting
- From Quantity to Quality: Boosting LLM Performance with Self-Guided Data Selection for Instruction Tuning
- TOFU: A Task of Fictitious Unlearning for LLMs
- Spherical Text Embedding
- Coresets for Data-efficient Training of Machine Learning Models
- Descent-to-Delete: Gradient-Based Methods for Machine Unlearning
- The Linear Representation Hypothesis and the Geometry of Large Language Models
- Active Learning for Convolutional Neural Networks: A Core-Set Approach
- Contrastive Representation Shaping for LLM Unlearning
- Gemma 2: Improving Open Language Models at a Practical Size
- Data Selection for Language Models via Importance Resampling
- Relative Density-Ratio Estimation for Robust Distribution Comparison
- Negative Preference Optimization: From Catastrophic Collapse to Effective Unlearning
Related papers
- MAVEN-T: Reinforced Heterogeneous Distillation for Real-Time Multi-Agent Trajectory Prediction
- Model Discovery Agent: LLM-assisted Bayesian experiment design for data-efficient discovery of mechanistic world models
- Agents' Last Exam
- The Clinician's Veto: Navigating Trust, Liability, and Uncertainty in Autonomous AI Prescribing
- MindHelper: Closed-Loop Embodied Mental-State Reasoning for Precision Intervention
- Incumbent Advantage: Brand Bias and Cognitive Manipulation Dynamics in LLM Recommendation Systems