Universal Reasoning Model
arxiv:2512.14693
Reasoning and the "physics" of language models · 2025-12-16
Complete cleaned Markdown · Source Markdown including bibliography
Read in parts
Contents
- Universal Reasoning Model
- Abstract
- 1 Introduction
- 2 Preliminaries
- 2.1 Standard Transformer
- 2.2 Universal Transformer
- 2.2.1 Parameter Sharing
- 2.2.2 Adaptive Computation Time (ACT)
- 3 Universal Reasoning Model
- 3.1 ConvSwiGLU
- 3.2 Truncated Backpropagation Through Loops
- 4 Experiment
- 4.1 Experiment Settings
- 4.2 Main Results
- 4.3 Why Universal Transformer?
- 4.4 Short Convolution
- 4.5 Truncated Backpropagation Through Loops
- 4.6 Nonlinearity of Transformers
- 4.7 Muon Optimizer
- 5 Related Work
- 5.1 ARC-AGI
- 5.2 Universal Transformers (Looped Transformers)
- 6 Conclusion
- 7 Acknowledgement