论文精选

Equilibrium Reasoners:学习吸引子实现可扩展推理

Equilibrium Reasoners: Learning Attractors Enables Scalable Reasoning

精选理由

EqR 用吸引子理论解释了为什么迭代推理能泛化,做推理模型或可扩展计算的团队值得关注——它可能改变你对测试时计算分配的理解。

Equilibrium Reasoners (EqR) 是一种新的推理框架,通过将推理过程建模为学习任务条件吸引子(latent dynamical systems)来实现可扩展的测试时计算。

正文

该框架无需外部验证器或任务特定先验,通过增加迭代深度(更多步数)和广度(聚合多个随机轨迹)来提升性能。实验表明,测试时计算的增益与向解对齐吸引子的收敛程度紧密相关。

在 Sudoku-Extreme 任务上,EqR 通过展开多达 40,000 层,将前馈模型的准确率从 2.6% 提升至超过 99%。

这一视角为理解迭代潜在模型中的可扩展推理提供了机制性解释。

原文 · arXiv cs.LG

Equilibrium Reasoners: Learning Attractors Enables Scalable Reasoning

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.

产业坐标与信源核验

本条情报归属 AITOP「论文」核心追踪栏目。涉及核心实体与技术议题包括:推理模型、测试时计算、吸引子、可扩展性,读者可通过下方标签进入对应专题页查阅历史脉络与关联对比。

信源核验说明:本文原始线索采编自「arXiv cs.LG」(@Benhao Huang, Zhengyang Geng, Zico Kolter),经 AITOP 多维评分管线结构化提炼。