AI Paper Atlas

Research map / Continual learning / Continual fine-tuning of LLMs

Forget to Know, Remember to Use: Context-Aware Unlearning for Large Language Models

Yuefeng Peng, Parnian Afshar, Megan Ganji, Thomas Butler, Amir Houmansadr, Mingxian Wang, Dezhi Hong

ICML 2026 · Continual learning › Continual fine-tuning of LLMs · approach: Machine unlearning by selective forgetting · 0 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
add KL context termRMU · TOFU 5% forget, Gemma-2B-ITContextual QA LLM-Judge: 0.00 -> 0.99Improved
add KL context termRMU · TOFU 5% forget, Qwen3-8BContextual QA LLM-Judge: 0.05 -> 0.97Improved
add KL context termNPO · TOFU 5% forget, Gemma-2B-ITContextual QA LLM-Judge: 0.81 -> 0.98Improved
add KL context termNPO · TOFU 5% forget, Qwen3-8BContextual QA LLM-Judge: 0.84 -> 0.95Improved
add KL context termUNDIAL · TOFU 5% forget, Gemma-2B-ITContextual QA LLM-Judge: 0.82 -> 0.98Improved
add KL context termUNDIAL · TOFU 5% forget, Qwen3-8BContextual QA ROUGE-L: 0.59 -> 0.68Improved
add KL context termRMU · TOFU 5% forget, Gemma-2B-ITDirect QA ROUGE-L (forgetting): 0.04 -> 0.13No clear effect
add KL context termNPO · TOFU 5% forget, Gemma-2B-ITModel utility: 0.57 -> 0.57No clear effect
add KL context termRMU · PISTOL A_C, Gemma-2B-ITContextual QA LLM-Judge: 0.00 -> 0.80Improved
add KL context termGradDiff · TOFU 5% forget, Gemma-2B-ITContextual QA LLM-Judge: 0.000 -> 0.945Improved
add KL context termSimNPO · TOFU 5% forget, Gemma-2B-ITContextual QA LLM-Judge: 0.940 -> 0.970Improved
add KL context termRMU · TOFU 5% forget, long noisy context, Gemma-2B-ITContextual QA LLM-Judge: 0.010 -> 0.605Improved

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