YANchor-4B: Effective Long-Horizon Reasoning in O(N) Time with O(1) Memory

Abstract

Long-horizon reasoning demands access to earlier information at a manageable generation cost. Full-history attention incurs growing storage and computation, while recurrent compression can lose precise details. Therefore, we present YANchor-4B, a general-purpose recurrent model that preserves crucial memory as ANchors for retrieval during subsequent reasoning. Beyond O(N)-time generation and O(1) memory, YANchor enables effective long-horizon reasoning through its multidimensional memory mechanism. For example, on challenging math problems, it achieves 82.93% mean pass@1 on AIME 2024--2026 and 63.64% on HMMT, substantially outperforming linear-time, constant-state counterparts, including larger models. It also delivers several-fold higher batched long-generation throughput than Transformer and hybrid baselines on H100. Furthermore, evaluations across dozens of benchmarks demonstrate YANchor's superiority in general-purpose capabilities.

Full paper on arXiv

Details

arXiv identifier

2610.10118

Published

2026-10-07

Authors

Huishan Ji, Hua Xu, Weiming Zhang, Qirui Ye

Open Models Built on This Paper

Every model in the SAVRN Model Hub whose card cites this paper, most downloaded first, with what it takes to run each one.