Looped Transformers with Source-Centered State Evolution
Bum Jun Kim, Kohei Hayashi, Shunsuke Kamiya, Masanori Koyama, Yusuke Iwasawa, Yutaka Matsuo
cs.LG, cs.CL
Submitted: 2026-07-30
Comments: 24 pages, 5 figures
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Terminology
Sources
- Layer Normalization
- PonderNet: Learning to Ponder
- A Mechanistic Analysis of Looped Reasoning Language Models
- Thinking Deeper, Not Longer: Memory-Efficient Test-Time Reasoning with Depth-Recurrent Transformers for Compositional Generalization
- Scaling up Test-Time Compute with Latent Reasoning: A Recurrent Depth Approach
- Adaptive Computation Time for Recurrent Neural Networks
- Stable Architectures for Deep Neural Networks
- Training Compute-Optimal Large Language Models
- LoopFormer: Elastic-Depth Looped Transformers for Latent Reasoning via Shortcut Modulation
- Scaling Laws for Neural Language Models
- Bridging the Gaps Between Residual Learning, Recurrent Neural Networks and Visual Cortex
- LoopUS: Recasting Pretrained LLMs into Looped Latent Refinement Models
- Parcae: Scaling Laws For Stable Looped Language Models
- Mixture-of-Depths: Dynamically allocating compute in transformer-based language models
- Dense Supervision Is Not Enough: The Readout Blind Spot in Looped Language Models
- Highway Networks
- Stabilizing Recurrent Dynamics for Test-Time Scalable Latent Reasoning in Looped Language Models
- Hyperloop Transformers
- 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