AI Paper Atlas

Research map / Large language models / Language, safety and interpretability

Beyond Superficial Forgetting: Thorough Unlearning Through Knowledge Density Estimation and Block Re-Insertion

Feng Guo, Yuntao Wen, Shen Gao, Junshuo Zhang, Shuo Shang

AAAI 2026 · Large language models › Language, safety and interpretability · approach: LLM unlearning and benchmarks

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add block re-insertion into original model (ablation)KUnBR · Random BirthdaysRTT forget accuracy (%): 80.2 -> 43.9Improved
add block re-insertion into original model (ablation)KUnBR · MMLURTT forget accuracy (%): 57.6 -> 28.0Improved
add block re-insertion into original model (ablation)KUnBR · YearsRTT forget accuracy (%): 68.3 -> 36.0Improved
add gradient-difference pre-unlearning warm-up (ablation)KUnBR · Random BirthdaysRTT forget accuracy (%): 54.1 -> 43.9Improved
add gradient-difference pre-unlearning warm-up (ablation)KUnBR · YearsRTT forget accuracy (%): 36.7 -> 36.0No clear effect
select blocks by knowledge density instead of head, bottom or uniform (ablation)KUnBR · Yearsforget accuracy over epochsImproved
use fewer blocks (M=4) instead of M=8 (ablation)KUnBR · YearsRTT forget accuracy (%) at Top-K=50%: 30.4 -> 45.6Hurt
use too many blocks (M=32) instead of M=8 (ablation)KUnBR · YearsRTT forget accuracy (%) at Top-K=75%: 15.2 -> 29.1Hurt
KUnBR vs best baseline on RTT robustnessKUnBR · WMDP-Deduped (LLaMA3-8B)RTT forget accuracy (%): 52.2 -> 38.8Improved
KUnBR vs NPO and RIA on general capabilityKUnBR · MMLUreasoning (Rea.): 0.00 -> 41.1Improved
KUnBR on Zephyr-7B recoveryKUnBR · Random Birthdays (Zephyr-7B-beta)recovery accuracy (%): 7.64 -> 20.4Hurt
KUnBR training time vs GDKUnBR · Yearsminutes: 20 -> 17Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy