AnalysisAI ModelsJuly 15, 2026

New papers propose improved methods for LLM unlearning

At least 7 Arxiv papers (June–July 2026) introduce techniques like signal-guided optimization, off-policy replay, and representation selectivity to improve LLM unlearning. Methods aim to balance forgetting specific knowledge while preserving general capabilities.

How this story unfolded

3 weeks · 7 reports · 7 of 14 shown

Jul 15
- Signal-Guided Optimization for Machine Unlearningarxiv.org
Jul 23
- Unlearning as Distribution Restoration: A Controlled Counterfactual Study, a Validated Selective Screen, and the Limits of Oracle-Free Certificationarxiv.org
Aug 3
- TextCloak: Thwarting Unauthorized LLM Exploitation via RL-Driven Unlearnable Textarxiv.org
- MMFGU: Multimodal Federated Graph Unlearningarxiv.org
Aug 4
- Leak It: A Probabilistic Approach to Training-Data Extraction from Black-Box Language Modelsarxiv.org
- Similarity-Aware Machine Unlearningarxiv.org
Aug 6
- Leak-Resistant Unlearning: A New Benchmark for Evaluating Multi-Hop Reasoning Consistency and Recovery Robustnessarxiv.org

Daily brief

Get tomorrow's AI brief in your inbox

New papers propose improved methods for LLM unlearning

How this story unfolded

Daily brief

More stories today