# Universal Reasoning Model

arXiv: https://arxiv.org/abs/2512.14693 · HTML: https://mlsanity.com/paper/2512.14693/

## Key facts

- Cited by 11 papers in the map (data through 14 Oct 2026).
- Placed in Reasoning benchmarks and evaluation, one of 17 directions in LLM reasoning.
- Builds on 13 of its 23 references that are in the map.
- First posted on arXiv 16 Dec 2025.

## Placed in

- LLM reasoning / [Reasoning benchmarks and evaluation](https://mlsanity.com/lr/reasoning-benchmarks-and-evaluation/)

## Builds on (most cited first, 13 of 13)

- [GLU Variants Improve Transformer](https://mlsanity.com/paper/2002.05202/) (2020)
- [Universal Transformers](https://mlsanity.com/paper/1807.03819/) (2018)
- [Adaptive Computation Time for Recurrent Neural Networks](https://mlsanity.com/paper/1603.08983/) (2016)
- [Muon is Scalable for LLM Training](https://mlsanity.com/paper/2502.16982/) (2025)
- [ARC-AGI-2: A New Challenge for Frontier AI Reasoning Systems](https://mlsanity.com/paper/2505.11831/) (2025)
- [Hierarchical Reasoning Model](https://mlsanity.com/paper/2506.21734/) (2025)
- [ARC Prize 2024: Technical Report](https://mlsanity.com/paper/2412.04604/) (2024)
- [Grokked Transformers are Implicit Reasoners: A Mechanistic Journey to the Edge of Generalization](https://mlsanity.com/paper/2405.15071/) (2024)
- [Less is More: Recursive Reasoning with Tiny Networks](https://mlsanity.com/paper/2510.04871/) (2025)
- [Unbiasing Truncated Backpropagation Through Time](https://mlsanity.com/paper/1705.08209/) (2017)
- [ARC Is a Vision Problem!](https://mlsanity.com/paper/2511.14761/) (2025)
- [Hierarchical Reasoning Models: Perspectives and Misconceptions](https://mlsanity.com/paper/2510.00355/) (2025)
- [What Makes Looped Transformers Perform Better Than Non-Recursive Ones](https://mlsanity.com/paper/2510.10089/) (2025)

## Built on by (most cited first, 11 of 11)

- [Fixed-Point Reasoners: Stable and Adaptive Deep Looped Transformers](https://mlsanity.com/paper/2606.18206/) (2026)
- [Step-resolved data attribution for looped transformers](https://mlsanity.com/paper/2602.10097/) (2026)
- [Universal Transformers Need Memory: Depth-State Trade-offs in Adaptive Recursive Reasoning](https://mlsanity.com/paper/2604.21999/) (2026)
- [One Step Forward and K Steps Back: Better Reasoning with Denoising Recursion Models](https://mlsanity.com/paper/2604.18839/) (2026)
- [Recursive Inference Machines for Neural Reasoning](https://mlsanity.com/paper/2603.05234/) (2026)
- [Steering Recurrent Reasoners at Inference Time with Readout Feedback](https://mlsanity.com/paper/2608.24136/) (2026)
- [LoopCD: Loop-wise Contrastive Decoding for Improving Reasoning in Looped Language Models](https://mlsanity.com/paper/2609.24196/) (2026)
- [Structured Sparse Memory for Recurrent Reasoning](https://mlsanity.com/paper/2609.33270/) (2026)
- [What Limits Recursive Reasoning Models: Optimization, Architecture and Test-Time Scaling](https://mlsanity.com/paper/2609.39967/) (2026)
- [Unrolled Flow Models for Reasoning](https://mlsanity.com/paper/2610.09759/) (2026)
- [Scaling to Tens of Thousands of Test-Time Iterations with Loop-Native Attention Residuals](https://mlsanity.com/paper/2610.11570/) (2026)

## Abstract (arXiv)

Universal transformers (UTs) have been widely used for complex reasoning tasks such as ARC-AGI and Sudoku, yet the specific sources of their performance gains remain underexplored. In this work, we systematically analyze UTs variants and show that improvements on ARC-AGI primarily arise from the recurrent inductive bias and strong nonlinear components of Transformer, rather than from elaborate architectural designs. Motivated by this finding, we propose the Universal Reasoning Model (URM), which enhances the UT with short convolution and truncated backpropagation. Our approach substantially improves reasoning performance, achieving state-of-the-art 53.8% pass@1 on ARC-AGI 1 and 16.0% pass@1 on ARC-AGI 2. Our code is avaliable at https://github.com/UbiquantAI/URM.