AI Paper Atlas

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

Revisiting the Past: Data Unlearning with Model State History

Keivan Rezaei, Mehrdad Saberi, Abhilasha Ravichander, Soheil Feizi

ICLR 2026 · Large language models › Language, safety and interpretability · approach: LLM pruning and scaling data · 1 citations (OpenAlex)

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
use checkpoint close to forget set (3691B/3859B/last vs 500B) (ablation)MSA · TOFU forget10 (OLMo-2-7B)Acc_retain: 0.64 -> 0.83Improved
apply MSA with pre-exposure checkpoint instead of Task Vector from target model (ablation)MSA · TOFU forget10 (OLMo-2-7B)Acc_recover: 0.26 -> 0.42Improved
MSA vs RMU on recovering ideal-model behaviorMSA · TOFU forget10 (OLMo-2-7B)Acc_recover: 0.08 -> 0.42Improved
MSA vs NPO on RESTORMSA_instruct · RESTOR Llama-3.1-8B-Instructaccuracy: 48.45 -> 63.95Improved
MSA vs best baseline on RESTOR OLMo-2-7BMSA_last · RESTOR OLMo-2-7Baccuracy: 40.25 -> 47.80Improved
use earlier checkpoint (500B instead of last) (ablation)MSA · RESTOR OLMo-2-7Baccuracy: 47.80 -> 45.67Mixed
use earlier checkpoint (500B instead of last) (ablation)MSA · MUSE-Books KnowMem retainKnowMem: 0.65 -> 0.51Hurt
MSA vs baselines on membership inferenceMSA_last · MUSE-BooksPrivLeak: 14.67 vs -99.93 (NPO), -17.83 (RMU)Improved
use checkpoint before forget-set exposure rather than checkpoint after TOFU when forget set is not last (ablation)MSA · TOFU forget10 + C4 (Llama-3.2-1B)Acc_recover: 0.37 -> 0.45Improved
base vs instruct checkpoint for forget vector (ablation)MSA · TOFU forget10 (Llama-3.1-8B-Instruct)Acc_recover: 0.45 -> 0.46No clear effect

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