# Speed is Confidence

arXiv: https://arxiv.org/abs/2601.19085 · HTML: https://mlsanity.com/paper/2601.19085/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 12 of its 28 references that are in the map.
- First posted on arXiv 27 Jan 2026.

## Builds on (most cited first, 12 of 12)

- [Outrageously Large Neural Networks: The Sparsely-Gated Mixture-of-Experts Layer](https://mlsanity.com/paper/1701.06538/) (2017)
- [GLU Variants Improve Transformer](https://mlsanity.com/paper/2002.05202/) (2020)
- [Generating Long Sequences with Sparse Transformers](https://mlsanity.com/paper/1904.10509/) (2019)
- [MLP-Mixer: An all-MLP Architecture for Vision](https://mlsanity.com/paper/2105.01601/) (2021)
- [Universal Transformers](https://mlsanity.com/paper/1807.03819/) (2018)
- [Adaptive Computation Time for Recurrent Neural Networks](https://mlsanity.com/paper/1603.08983/) (2016)
- [PonderNet: Learning to Ponder](https://mlsanity.com/paper/2107.05407/) (2021)
- [T2FSNN: Deep Spiking Neural Networks with Time-to-first-spike Coding](https://mlsanity.com/paper/2003.11741/) (2020)
- [PolyNet: Learning Diverse Solution Strategies for Neural Combinatorial Optimization](https://mlsanity.com/paper/2402.14048/) (2024)
- [DESlib: A Dynamic ensemble selection library in Python](https://mlsanity.com/paper/1802.04967/) (2018)
- [Window-Based Early-Exit Cascades for Uncertainty Estimation: When Deep Ensembles are More Efficient than Single Models](https://mlsanity.com/paper/2303.08010/) (2023)
- [Latent Reasoning in TRMs is Secretly a Policy Improvement Operator](https://mlsanity.com/paper/2511.16886/) (2025)

## Built on by (most cited first, 1 of 1)

- [Resample the Answer: Where to Spend Test-Time Compute in Recursive Reasoning](https://mlsanity.com/paper/2610.16427/) (2026)

## Abstract (arXiv)

Biological neural systems must be fast but are energy-constrained. Evolution's solution: act on the first signal. Winner-take-all circuits and time-to-first-spike coding implicitly treat when a neuron fires as an expression of confidence. We apply this principle to ensembles of Tiny Recursive Models (TRM) [Jolicoeur-Martineau et al., 2025]. On Sudoku-Extreme, halt-first selection achieves 97% accuracy vs. 91% for probability averaging -- while requiring 10x fewer reasoning steps. A single baseline model achieves 85.5% +/- 1.3%. Can we internalize this as a training-only cost? Yes: by maintaining K=4 parallel latent states but backpropping only through the lowest-loss "winner," we achieve 96.9% +/- 0.6% accuracy -- matching ensemble performance at 1x inference cost, with less than half the variance of the baseline. A key diagnostic: 89% of baseline failures are selection problems, revealing a 99% accuracy ceiling. As in nature, this work was also resource constrained: all experiments used a single RTX 5090. A modified SwiGLU [Shazeer, 2020] made Muon [Jordan et al., 2024] and high LR viable, enabling baseline training in 48 minutes and full WTA (K=4) in 6 hours on consumer hardware.