Memorization in Language Models

Follow this topic Ask about this topic

Momentum

10 papers in the last four weeks, up 150% on the four weeks before. 0.1% of all new papers.

Jul 13Week of Sep 28

Latest papers 75

All topics

Probe-Geometry Alignment: Erasing the Cross-Sequence Memorization Signature Below Chance
May 3, 2026Anamika Paul Rupa, Anietie AndyMemorization in Language Models Neural Network Memorization
Howard University
Learning How and What to Memorize: Cognition-Inspired Two-Stage Optimization for Evolving Memory
May 1, 2026Derong Xu, Shuochen Liu, Pengfei Luo +8Memorization in Language Models Memorization in Generative Models
University of Science and Technology of China & State Key Laboratory of Cognitive Intelligence · City University of Hong Kong · Dalian University of Technology +1
Language Diffusion Models are Associative Memories Capable of Retrieving Unseen Data
Apr 29, 2026Bao Pham, Mohammed J. Zaki, Luca Ambrogioni +2Memorization in Language Models Neural Network Memorization
1Rensselaer Polytechnic Institute (RPI) · 2Donders Institute for Brain, Cognition, and Behaviour, Radboud University · 3Dynamical Mind +1
ZenBrain: A Neuroscience-Inspired 7-Layer Memory Architecture for Autonomous AI Systems
Apr 26, 2026Alexander BeringMemorization in Language Models Neural Processes
Zensation AI
Revisiting Non-Verbatim Memorization in Large Language Models: The Role of Entity Surface Forms
Apr 23, 2026Yuto Nishida, Naoki Shikoda, Yosuke Kishinami +4LLM Evaluation Memorization in Language Models
Nara Institute of Science and Technology · Future Corporation
EngramaBench: Evaluating Long-Term Conversational Memory with Structured Graph Retrieval
Apr 23, 2026Julian AcunaLLM Evaluation Memorization in Language Models
SCM: Sleep-Consolidated Memory with Algorithmic Forgetting for Large Language Models
Apr 22, 2026Saish Sachin ShindeMemory-Augmented Language Models Memorization in Language Models
Clyrai IP Studio
Understanding Secret Leakage Risks in Code LLMs: A Tokenization Perspective
Apr 20, 2026Meifang Chen, Zhe Yang, Huang Nianchen +4Data Leakage Memorization in Language Models
†The Chinese University of Hong Kong · ‡Nanyang Technological University · §University of Southern California +1
AnchorMem: Anchored Facts with Associative Contexts for Building Memory in Large Language Models
Apr 19, 2026Zhanyu Shen, Sijie Cheng, Zhicheng Guo +3Memory-Augmented Language Models Memorization in Language Models
College of Computer Science and Software Engineering, Shenzhen University · Tsinghua University
HeLa-Mem: Hebbian Learning and Associative Memory for LLM Agents
Apr 18, 2026Jinchang Zhu, Jindong Li, Cheng Zhang +2Memorization in Language Models Hebbian Learning
The Hong Kong University of Science and Technology (Guangzhou) · Jilin University · The Chinese University of Hong Kong
Alignment Whack-a-Mole : Finetuning Activates Verbatim Recall of Copyrighted Books in Large Language Models
Mar 21, 2026Xinyue Liu, Niloofar Mireshghallah, Jane C. Ginsburg +1LLM Alignment Memorization in Language Models
Stony Brook University · Carnegie Mellon University · Columbia Law School
Spurious Rewards Paradox: Mechanistically Understanding How RLVR Activates Memorization Shortcuts in LLMs
Jan 16, 2026Lecheng Yan, Ruizhe Li, Guanhua Chen +5Reward Hacking Memorization in Language Models
Southern University of Science and Technology · University of Science and Technology of China · University of Aberdeen +5
MemBuilder: Reinforcing LLMs for Long-Term Memory Construction via Attributed Dense Rewards
Jan 9, 2026Zhiyu Shen, Ziming Wu, Fuming Lai +2Reinforcement Learning Memory-Augmented Language Models
School of Computer Science and Engineering, Sun Yat-sen University, Guangzhou, China · Tencent Inc.
Continual Memorization of Factoids in Language Models
Nov 11, 2024Howard Chen, Jiayi Geng, Adithya Bhaskar +2Continual Learning for LLMs Memorization in Language Models
Princeton Language and Intelligence (PLI), Princeton University
Mitigating Memorization In Language Models
Oct 3, 2024Mansi Sakarvadia, Aswathy Ajith, Arham Khan +6Memorization in Language Models LLM Fine-Tuning
University of Chicago · Argonne National Laboratory · Lawrence Berkeley National Laboratory +3

May 3, 2026, cs.LGJ/K move · Enter open · S save

Probe-Geometry Alignment: Erasing the Cross-Sequence Memorization Signature Below Chance

Anamika Paul Rupa, Anietie Andy

Howard University

Memorization in Language Models Neural Network Memorization Persistent Memory for Language Models LLM Memory Privacy Leakage in Language Models Memorization in Generative Models+4

Open paper AskPDF arXiv

Recent attacks show that behavioural unlearning of large language models leaves internal traces recoverable by adversarial probes. We characterise where this retention lives and show it can be surgically removed without measurable capability cost. Our central protocol is a leave-one-out cross-sequence probe that tests whether a memorisation signature generalises across held-out sequences. The signature is real and consistent across scale: memorisation-specific gaps of +0.32, +0.19, +0.30 on Pythia-70M, GPT-2 medium, and Mistral-7B; on Pythia-70M, the random-initialisation control collapses to -0.04 at the deepest layer where the pretrained signature peaks. The probe direction is causally separable from recall -- projecting it out collapses the signature locally (+0.44 -> -0.19) while behavioural recall barely changes -- and a probe trained on naturally memorised content does not classify fine-tuning-injected secrets, marking two representationally distinct regimes. We then introduce probe-geometry alignment (PGA), a surgical erasure that aligns activations along the probe's live readout direction at each depth. PGA drives the cross-sequence probe below random chance at all four scales tested (toy depth-4: 0.17; Pythia-70M: 0.07; Mistral-7B: 0.45; GPT-2 medium: 0.06 via MD-PGA k=2) and remains robust to six adversarial probe variants. Against a re-fitting attacker who trains a fresh probe on PGA-treated activations, we extend PGA adversarially, defeating the re-fit probe at every memorisation-relevant depth while preserving five zero-shot capability benchmarks within 2.8 percentage points per task (mean Δacc = +0.2pp). The cross-sequence signature is a real, causally separable, regime-specific property of pretrained representations -- removable below chance with a single rank-one intervention per depth at no measurable capability cost.

Momentum

Related topics

Latest papers 75