AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

Is the Reversal Curse a Binding Problem? Uncovering Limitations of Transformers from a Basic Generalization Failure

Boshi Wang, Huan Sun

ICLR 2026 · Architectures and efficiency › Attention and transformers · approach: Transformer expressivity and learnability

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
Represent inputs at the abstract concept level (own embedding per entity) instead of surface names (ablation)Standard decoder-only transformer · Reversal taskMRR: 0.0% accuracy (surface) -> MRR 0.951 (concept level, |E_A|=50K)Improved
Increase |E_A| (number of training entities) (ablation)Standard transformer, concept level · Abstract reversalMRR: 0.823 -> 0.964 (1 layer, 2.5K -> 100K)Improved
Use JEPA with in-batch contrastive loss at concept levelAutoregressive transformer · Surface-level reversalMRR: 0.0% -> 0.864Improved
Increase depth of recognition module or semantic-to-recognition ratio in JEPA (ablation)JEPA · Surface-level reversalMRR: 0.864 -> 0.230Hurt
Increase name multiplicity (more entities sharing tokens) (ablation)JEPA · Surface-level reversalMRRHurt
Increase model width from 768 to 1280 (ablation)JEPA · Surface-level reversalMRRImproved
Replace last MLP of recognition module with a memory layer (ablation)JEPA · Surface-level reversalMRRImproved
Add memory layers to a standard transformer without JEPA (ablation)Standard transformer · Surface-level reversalAccuracyNo clear effect
Vary depth or tied vs untied embeddings in standard surface-level training (ablation)Standard transformer · Surface-level reversalAccuracy (%): 0.0%No clear effect
Use parametric memory with JEPA and memory layers vs frontier LLMs with in-context factsLarge-scale arithmetic reasoning · Branching factor 10 to 40AccuracyImproved

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