The Surprising Effectiveness of Shared Memory in Looped Transformers

arXiv:2610.02383 · cs.LG, cs.AI · Submitted 2026-10-01 · Read on arXiv

cs.LG, cs.AI

Submitted: 2026-10-01

Updated: 2026-10-01

Code: https://github.com/karpathy/nanoGPT

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Terminology

• Transformer — 1× in this paper · explained in 2 episodes

Sources

• CART: Context-Anchored Recurrent Transformer -- A Parameter-Efficient Architecture with Learned Stability
• Post-LayerNorm Is Back: Stable, ExpressivE, and Deep
• Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challenge
• DeepSeek-V4.1-Flash: Pushing the Limits of KV Cache Compression
• Loop the Loopies!
• The State-Prediction Separation Hypothesis
• Recirculation
• Looped Latent Attention: Cross-Loop KV Compression for Looped Transformers
• GLU Variants Improve Transformer
• Universal YOCO for Efficient Depth Scaling
• SMELT: Scaling Laws for Compute-Matched MoE Looped Transformers
• Parallel Loop Transformer for Efficient Test-Time Computation Scaling
• Scaling Latent Reasoning via Looped Language Models

Related papers

• Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
• AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
• Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
• Convergence issues in Relational Concept Analysis based on AOC-posets
• Beliefs Beyond Posteriors: The Predictive Cost of ELBO Binding
• Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks