YANchor-4B: Effective Long-Horizon Reasoning in O(N) Time with O(1) Memory
Therefore, we present YANchor-4B, a general-purpose recurrent model that preserves crucial memory as ANchors for retrieval during subsequent reasoning.
Key points
- 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.
- 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.
Sources (1)
- [1]YANchor-4B: Effective Long-Horizon Reasoning in O(N) Time with O(1) MemoryarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 7, 02:03 PM
Therefore, we present YANchor-4B, a general-purpose recurrent model that preserves crucial memory as ANchors for retrieval during subsequent reasoning.
Long-horizon reasoning demands access to earlier information at a manageable generation cost.
Extractive summary: sentences quoted from the sources.