AI Paper Atlas

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

Dual-Space Smoothness for Robust and Balanced LLM Unlearning

Han Yan, Zheyuan Liu, Meng Jiang

ICLR 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 parameter-space smoothness (ablation)PRISM · MUSE-Books relearning attack 100 stepsVerbMem on forget set: 16.664 -> 6.804Improved
add representation-space smoothness (robust probe) (ablation)PRISM · MUSE-Books relearning attack 100 stepsVerbMem on forget set: 11.112 -> 6.804Improved
add gradient-conflict decoupling (ablation)PRISM · MUSE-Books relearning attack 50 stepsutility on retain set: 1.333 -> 46.588Improved
add representation-space smoothing (ablation)PRISM · MUSE-Books relearning attack 50 stepsutility on retain set: 24.412 -> 46.588Improved
add parameter-space smoothing (ablation)PRISM · MUSE-Books relearning attack 50 stepsutility on retain set: 13.506 -> 46.588Improved
use PRISM instead of SAM+NPOPRISM · MUSE-NewsUnlearn Score: 0.000 -> 0.522Improved
use PRISM instead of RMU-LAT under Relearn-25% at 50 stepsPRISM · MUSE-Books Relearn-25%VerbMem on forget set: 93.87 -> 0.082Improved
add PRISM to the unlearned modelPRISM · MUSE-Booksrepresentation margin: median margin +24.9%Improved
use PRISM instead of GDR-regularized baselinesPRISM · WMDP XStestrefusal rate: 0.880 -> 1.000Hurt
use PRISM instead of SAM+NPOPRISM · MUSE-Bookstime per step (s): 11.055 -> 11.223Hurt
use PRISM instead of NPO_GDRPRISM · WMDP multi-turn jailbreakASR: 0.326 -> 0.196Improved

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