AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

AI Engram: In Search of Memory Traces in Artificial Intelligence

Jea Kwon, Dong-Kyum Kim, Jiwon Kim, Yonghyun Kim, Woong Kook, Meeyoung Cha

ICML 2026 · Architectures and efficiency › RNNs and normalization · approach: Brain-inspired network comparison · 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 Engram (alpha=1) to unlearn class 0ResNet-18 · CIFAR-10 class-0 unlearningDA: 0.987 (retrain) vs 0.992Improved
use alpha best instead of alpha=1 (ablation)Engram, CIFAR-10 ResNet-18 · CIFAR-10 class-0 unlearningToW: 0.930 -> 0.984Improved
use alpha best instead of alpha=1 (ablation)Engram, CIFAR-10 ResNet-18 · CIFAR-10 class-0 unlearningNMI: 0.379 -> 0.611Hurt
use Engram on CIFAR-100ResNet-50 · CIFAR-100 class-0 unlearningToW: 0.988 (alpha=1) vs retrain 0.985Improved
use adaptive W-Norm alpha instead of uniform alpha=0.6 (ablation)Engram, Llama-3.2-1B · TOFU forget10memorization: 0.9176 -> 0.9627Improved
use adaptive W-Norm alpha instead of uniform alpha=0.6 (ablation)Engram, Llama-3.2-1B · TOFU forget10utility: 0.8801 -> 0.9256Improved
use Engram instead of UCELlama-3.2-1B · TOFU forget10Overall: 0.659 -> 0.818Improved
use Engram instead of iterative methods (AltPO, NPO, SimNPO)Llama-3.2-1B · TOFU forget10Overall / PrivacyHurt
use Engram instead of NPO for exact memorizationLlama-3.2-1B · TOFU forget10exact memorization: 0.0948 -> 0.0276Improved
use tabula rasa W_ft instead of naive delta W_ft - W_pt (ablation)Engram, Llama-3.2-1B · TOFU forget10Overall: 0.446 -> 0.818Improved
use tabula rasa W_ft instead of naive delta (ablation)Engram, Llama-3.2-1B · TOFU forget10exact memorization: 0.617 -> 0.028Improved
restrict extraction to Q/K + Gate layers (ablation)Engram, Llama-3.2-1B · TOFU forget10OverallNo 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