AI Paper Atlas

Research map / Graph learning / Knowledge and temporal graphs

Bilinear relational structure fixes reversal curse and enables consistent model editing

Dong-Kyum Kim, Minsung Kim, Jea Kwon, Nakyeong Yang, Meeyoung Cha

ICLR 2026 · Graph learning › Knowledge and temporal graphs · approach: Knowledge graph embeddings

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
increase weight decay to at least 1.0 during training (ablation)GPT-NeoX 206M trained from scratch on family KG · Unseen father/mother testtest accuracy: low accuracy (weight decay < 1.0) -> near-perfect accuracy for some modelsImproved
bilinear probe instead of linear or translational probe (ablation)Not Reversal Cursed model · Layer-wise probe accuracyprobe accuracy: above 95% vs near or below 1/3 baselineImproved
increase linear probe samples n from 10 to 500 (ablation)Linear relational embedding probe · Probe accuracyprobe accuracyNo clear effect
increase bilinear probe samples n (insufficient n) (ablation)Bilinear probe · Probe accuracyprobe accuracyImproved
model has bilinear structure (Not Reversal Cursed vs Reversal Cursed)Fine-tuning model edit · Logical generalization after editlogical generalizationImproved
model has bilinear structureFine-tuning model edit · LocalitylocalityImproved
model has bilinear structureFine-tuning model edit · Direct edit successedit successNo clear effect
edit early-to-mid layers (1-4) rather than layers 6-9 (ablation)Not Reversal Cursed models · Logical generalization after editlogical generalization: layers 1-4 best vs bilinear peak at layers 6-9Improved

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