AI Paper Atlas

Research map / Large language models / Retrieval and knowledge

Editing as Unlearning: Are Knowledge Editing Methods Strong Baselines for Large Language Model Unlearning?

Zexi Li, Xiangzhu Wang, William F. Shen, Meghdad Kurmanji, Xinchi Qiu, Dongqi Cai, Chao Wu, Nicholas D. Lane

AAAI 2026 · Large language models › Retrieval and knowledge · approach: Knowledge editing in LLMs · 10 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 editing methods (WISE, AlphaEdit) instead of unlearning baselines for pretrained knowledgeLLM unlearning · TOFU factual, Llama2-7B-chat1 - Rouge1@Forget + Rouge1@Retain: GD 1.321, DPO 1.086 -> WISE 1.481, AlphaEdit 1.606Improved
use editing methods instead of unlearning baselines for finetuned knowledgeLLM unlearning · PISTOL, Mistral-7B-instruct1 - Rouge1@Forget + Rouge1@Retain: AlphaEdit 1.074, WISE 1.309 vs GD 1.544, GA 1.493Mixed
add self-improvement pipeline to WISE and AlphaEdit (ablation)WISE, AlphaEdit · TOFU factual, Llama2-7B-chatEntailment (human rating): AlphaEdit 1.73 -> 2.12Improved
add self-improvement pipeline to WISE (ablation)WISE · TOFU factual, Llama2-7B-chatTrustworthiness (human rating): slight decline (-0.230)Hurt
add self-improvement pipeline to AlphaEdit (ablation)AlphaEdit · TOFU factual, Llama2-7B-chatRephrase (1 - Rouge1): 0.68 -> 0.90Improved
add query merging to ROME and MEMIT (ablation)ROME, MEMIT · TOFU factual, 80 forget samples, Llama2-7B-chat1 - Rouge1@Forget + Rouge1@Retain: ROME/MEMIT 1.622/1.632 at 5 merged queries, vs DPO 1.123, GD 1.596Improved
merge more queries into one (increase n) (ablation)ROME, MEMIT with query merging · TOFU factual, 80 forget samples1 - Rouge1@Forget + Rouge1@Retain: peak at 5 merged queriesMixed
increase number of forget samples from 20 to 100 (ablation)WISE, AlphaEdit · TOFU factual, Llama2-7B-chat1 - Rouge1 on forget setImproved
rephrase the forget queries (generalization attack) (ablation)all methods · TOFU factual, 40 forget samples, Llama2-7B-chatForget Rouge1: AlphaEdit 0.08 -> 0.14; WISE 0.28 -> 0.46Hurt
use ROME or MEMIT directly for unlearning without query mergingROME, MEMIT · TOFU factual, Llama2-7B-chatRetain Rouge1: MEMIT retain 0.01, ROME 0.04Hurt
use GRACE for unlearningGRACE · TOFU factual, Llama2-7B-chatForget Rouge1: 0.65 forget Rouge1 (0.82 before)No 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