Newest papers

MemT-1.0: Towards Memory-Native Evolution

Zhiyu Li, Zeyu Zhang, Linyun Liu, Yiwei Bu, Chenyang Xi, Xichong Zhang, Yi Yu, Yu Yu, Zhichen Liu, Yuxiao Yi, Chunyu Li, Haoran Liu, Zhenkai Ding, Le Zhou, Juncheng Zhang, Boyu Chen, Zehao Lin, Guorui Li, Renyu Fu, Ji Li · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.26825 · cited by 0 works (OpenAlex) · open-access PDF

German

Zu diesem Zweck führen wir MemT-1.0 ein, eine neue Architektur von Memory-Fundament-Modellen, die das Gedächtnis als Teil der Modellrechnung internalisieren. Aufbauend auf Metis nutzt MemT-1.0 Gated DeltaNet-2 mit einem stabilen Token-Level-Memory-Update-Mechanismus innerhalb jeder Runde. Wir führen weiterhin Mixture-of-Storage ein, um die native Memory-Kapazität zu skalieren. Nach einer bewussten Rückenlehnenvorbereitung entwickeln wir eine zweistufige Trainingspipeline, die sich aus kontinuierlichem Vor- und Nachtraining zusammensetzt, unterstützt von 34B-Token von Trainingsaufgaben und effizienter Ausbildung und Inferenzinfrastruktur. Wir veröffentlichen ein Familienmodell, das MemT

Machine translation (Meta M2M100); the English original follows.

Abstract

Memory-native evolution requires making memory an intrinsic capability of foundation models, enabling them to continuously accumulate and utilize information. To this end, we introduce MemT-1.0, a new architecture of memory foundation models that internalize memory as part of model computation. Building on Metis, MemT-1.0 employs Gated DeltaNet-2 with a stable token-level memory update mechanism within each turn. We further introduce Mixture-of-Storage to scale native memory capacity. Following a deliberate backbone preparation, we develop a two-stage training pipeline consisting of continued pre-training and post-training, supported by 34B tokens of training data and efficient training and inference infrastructure. We release a model family comprising MemT-1.0-Nano and MemT-1.0-Flash, and conduct a comprehensive evaluation of memory and general capabilities. The results show that they outperform the parametric memory baselines on average and remain competitive with retrieval-augmented methods, while largely preserving general capabilities. At a history length of 128K tokens, MemT-1.0-Nano achieves approximately $1.86\times$ end-to-end inference speedup and a $66\times$ reduction in persistent storage relative to full context KV cache. By internalizing retrieved procedural experience, MemT-1.0 improves task success on agentic tasks, providing a practical step toward memory-native evolution.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Computer science, Inference, Task (project management), Pipeline (software), Context (archaeology), Memory model

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange