Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
YANchor-4B is a general-purpose recurrent model designed for efficient long-horizon reasoning by employing a multidimensional memory mechanism that preserves crucial information as retrievable "Anchors." This approach achieves $O(N)$-time generation and $O(1)$ memory, significantly outperforming linear-time, constant-state models on challenging math benchmarks like AIME and HMMT. The model also demonstrates superior batched long-generation throughput compared to Transformer and hybrid baselines on H100, indicating its effectiveness across a broad range of general-purpose capabilities.
Long-horizon reasoning models can now achieve $O(1)$ memory and $O(N)$ time complexity while surpassing larger Transformer baselines on complex tasks.
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.