| title | Equilibrium Reasoners: Learning Attractors Enables Scalable Reasoning |
|---|---|
| authors | Benhao Huang, Zhengyang Geng, Zico Kolter |
| date | 2026-05-20 |
| venue | arXiv (preprint, not peer reviewed) |
| id | arXiv:2605.21488 |
| affiliation | Locus Lab |
| upvotes | 5 (Hugging Face Daily Papers, 2026-11-01) |
| links | abs pdf hf code |
| abstract | Scaling test-time compute by iteratively updating a latent state has emerged as a powerful paradigm for reasoning. Yet the internal mechanisms that enable these iterative models to generalize beyond memorized patterns remain unclear. We hypothesize that generalizable reasoning arises from learning task-conditioned attractors: latent dynamical systems whose stable fixed points correspond to valid solutions. We formalize this process through Equilibrium Reasoners (EqR), which enable test-time scaling without external verifiers or task-specific priors. EqR scales internal dynamics along two axes: depth, by running more iterations, and breadth, by aggregating stochastic trajectories from multiple initializations. Empirically, gains from test-time scaling are tightly coupled with stronger convergence toward solution-aligned attractors. This attractor perspective allows neural networks to adaptively allocate test-time compute based on task difficulty. While simple cases converge within 1 to 5 iteration steps, harder cases benefit from massive test-time scaling. By unrolling up to the equivalent of 40,000 layers, scalable latent reasoning boosts accuracy from 2.6% for feedforward models to over 99% on Sudoku-Extreme. These results suggest that learned attractor landscapes provide a useful mechanistic lens for understanding scalable reasoning in iterative latent models. |
cite
@misc{huang2026equilibrium,
title={Equilibrium Reasoners: Learning Attractors Enables Scalable Reasoning},
author={Benhao Huang and Zhengyang Geng and Zico Kolter},
year={2026},
eprint={2605.21488},
archivePrefix={arXiv},
url={https://arxiv.org/abs/2605.21488}
}