task · first seen 4 Oct 2026
LLM unlearning
Also: data deletion, exact unlearning, right to erasure, machine unlearning, selective forgetting, selective unlearning, right to be forgotten
Removing specific learned knowledge from a trained language model while preserving its other capabilities.
LLM unlearning, and machine unlearning more broadly, means removing specific data, knowledge or behaviors from a trained model without full retraining, often motivated by GDPR-style erasure requests or hazardous knowledge. The sources are almost all arXiv papers spread across many groups. They cover spectral and null-space projection updates for LLMs, exact unlearning via quantized sufficient statistics, role and persona unlearning with mixture-of-experts, tamper resistance through suppression during pretraining (an EPFL paper), and attacks that infer which concepts were erased from diffusion models. Unlearning is also being applied to face recognition, VLA robots, multi-agent RL and federated backdoor detection. One hackathon repo audits real-world data deletion.
This is a mature research area with steady output rather than a new spike. The open problems are robustness to fine-tuning and relearning, overlap between knowledge representations, and verifying that forgetting actually happened, and these will decide whether unlearning can meet regulatory or safety requirements.
Analyst note (established), 17 Nov 2026.
Mentions per week
PapersCodeDiscussion
Rank 133 on 14 Nov 2026 · 50 mentions in 28 days · 48 independent groups · papers → code · first seen 6 weeks ago · acceleration +0.03
Earliest sources
- Your Unlearning Gives You Away: Identifying Erased Concepts in Diffusion ModelsPapers · 4 Oct 2026 · Kaiyuan Deng, Yuchen Li, Yang Xiao, Bo Hui, Geng Yuan, Xiaolong Ma
- Exact Unlearning via Quantized Sufficient StatisticsPapers · 5 Oct 2026 · Ami Tavory, Shripad Gade, Tal Sarig, Noam Touitou, Ido Guy
- How Learning Governs Unlearning across the Memorization-Generalization SpectrumPapers · 6 Oct 2026 · Hwiyeong Lee, Hyelim Lim, Ingyu Bang, Hoki Kim, Taeuk Kim
- Sparse Feature Policy Unlearning Mitigates State Hallucination in Vision-Language-Action ModelsPapers · 7 Oct 2026 · Jiho Lee, Jeongeun Park, Heayoun Choi, Taekyung Kim, Eunwoo Kim
- Nullify: Null-Space Activation Steering for Training-Free LLM UnlearningPapers · 7 Oct 2026 · Wei Zhai, Xiang Liu, Qiang Huang, Rui Qian, Lemao Liu, Ziwei Li, Ziqi Wang, Zhit…
- Open-MMUnlearning: Unifying Methods and Evaluation for MLLM UnlearningPapers · 7 Oct 2026 · Junkai Chen, Yuhao He, Qianshan Wei, Junxiang You, Jingwen Shao, Junkai Lin, Zho…
Latest sources
- LEDGER: Auditable Agent Memory Under ChangePapers · 16 Nov 2026 · Md Aidul Islam, Md Nasir Uddin Shuvo, Muhammad Waseem, Pekka Abrahamsson
- LEAF-OPD: Learning to Forget from Privileged Contexts via On-Policy DistillationPapers · 16 Nov 2026 · Zirui Pang, Zhijie Deng, Yuancheng Wang, Qihan Yu, Xin Li, Jiaheng Wei, Zixin Zh…
- Counterfactual Feature Steering for Adjacency-Preserving Identity UnlearningPapers · 16 Nov 2026 · FNU Shivam, Layth Sliman, Ramana Reddy, Donald Adjeroh, Binod Bhattarai, Prashnn…
- Model Immunization Beyond Condition Numbers: The Importance of Plateau RegionsPapers · 16 Nov 2026 · Quoc Nguyen, Trung Le, Dinh Phung, Mehrtash Harandi
- From Facts to Personas: Interpretable Role Unlearning in LLMs via Mixture-of-ExpertsPapers · 14 Nov 2026 · Ruihong Zeng, Puning Yang, Jinghui Zhang, Shen Gao, Xiangliang Zhang, Preslav Na…
- When Cooperation Must Change: Semantic Strategy Unlearning in Multi-Agent Reinforcement LearningPapers · 13 Nov 2026 · Wenhan Wu, Linghan Chen, Yudong Gao, Fan Yang, Honglong Chen, Kai Yin, Jiawei Ji…
- Disentangling Forgetting from Utility: Lightweight LLM Unlearning via Orthogonal Null-Space ProjectionPapers · 13 Nov 2026 · Xiuyuan Wang, Weiming Liu, Zongxin Yang, Fan Wang, Liemiao Jin, Rubin Qin, Wei C…
- ShadowBench: Exposing Lexical Anchoring and the Illusion of Forgetting in Large Language ModelsPapers · 11 Nov 2026 · Sujan Maharjan, Lu Liu
- DETER: Suppression During Pretraining for Tamper ResistancePapers · 11 Nov 2026 · Maria Ios Glarou, Martin Jaggi
- Protecting Sensitive Knowledge Under LLM Unlearning: The Proxy Threat and a Superposition-Aware DefensePapers · 11 Nov 2026 · Jiayang Li, Tianqing Zhu, Longxiang Gao, Huajie Chen, Dayong Ye, Shang Wang, Hua…
- Data-Efficient Unlearning-Based Backdoor Detection in Federated LearningPapers · 10 Nov 2026 · Dowon Kim, Junhao Cai, Sung Il Choi, Dohun Kim, Juhyun Park, Chengjun Jin, Chang…
- Spectral Unlearning: Transformer Structure-Preserving Updates for Language ModelsPapers · 9 Nov 2026 · Sung Il Choi, Junhao Cai, Dohun Kim, Changhee Joo