Changsheng Wang

Downgrade to Upgrade: Optimizer Simplification Enhances Robustness in LLM Unlearning

LLM Unlearning Under the Microscope: A Full-Stack View on Methods and Metrics

The Fragile Truth of Saliency: Improving LLM Input Attribution via Attention Bias Optimization

Invariance Makes LLM Unlearning Resilient Even to Unanticipated Downstream Fine-Tuning

LLM Unlearning Reveals a Stronger-Than-Expected Coreset Effect in Current Benchmarks