Contrastive Representation Shaping for LLM Unlearning

arXiv:2601.22028 · cs.LG · Submitted 2026-01-29 · Read on arXiv

cs.LG

Submitted: 2026-01-29

Updated: 2026-09-30

Code: https://github.com/HaoranTang/CLReg

Terminology

• Entanglement — 2× in this paper · explained in 4 episodes
• Alignment — 1× in this paper · explained in 1 episode
• Benchmark — 1× in this paper · explained in 2 episodes
• Distribution Shift — 1× in this paper · explained in 7 episodes

Sources

• A Theoretical Analysis of Contrastive Unsupervised Representation Learning
• SalUn: Empowering Machine Unlearning via Gradient-based Weight Saliency in Both Image Classification and Generation
• Unraveling Meta-Learning: Understanding Feature Representations for Few-Shot Tasks
• The Llama 3 Herd of Models
• Approximate Data Deletion from Machine Learning Models
• Towards Unbounded Machine Unlearning
• LLM Unlearning with LLM Beliefs
• The WMDP Benchmark: Measuring and Reducing Malicious Use With Unlearning
• TOFU: A Task of Fictitious Unlearning for LLMs
• UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction
• SimPO: Simple Preference Optimization with a Reference-Free Reward
• Representation Learning with Contrastive Predictive Coding
• Direct Preference Optimization: Your Language Model is Secretly a Reward Model
• MUSE: Machine Unlearning Six-Way Evaluation for Language Models
• Sharpness-Aware Machine Unlearning
• Llama 2: Open Foundation and Fine-Tuned Chat Models
• Understanding Contrastive Representation Learning through Alignment and Uniformity on the Hypersphere
• Contrastive Preference Optimization: Pushing the Boundaries of LLM Performance in Machine Translation
• Negative Preference Optimization: From Catastrophic Collapse to Effective Unlearning
• What makes unlearning hard and what to do about it

Related papers

• Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
• AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
• Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
• Convergence issues in Relational Concept Analysis based on AOC-posets
• Beliefs Beyond Posteriors: The Predictive Cost of ELBO Binding
• Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks