#machine learning
Feb 2026
GONE: Structural Knowledge Unlearning via Neighborhood-Expanded Distribution Shaping
Graph Oblivion and Node Erasure (GONE), a benchmark for evaluating knowledge unlearning over structured knowledge graph (KG) facts in LLMs, enables the disentanglement of three effects of unlearning: direct fact removal, reasoning-based leakage, and catastrophic forgetting.
Chahana Dahal, A. Balasubramaniam, Zuo-Bin Xiong
· arXiv.org · 1 citation