The Memory-Augmented Model (M+) is a memory-augmented model based on MemoryLLM proposed by the University of California, San Diego and multiple research teams on May 30, 2025. It can significantly improve the ability to retain long-term information. The relevant paper results are "M+: Extending MemoryLLM with Scalable Long-Term Memory".

M+ integrates a long-term memory mechanism with a co-trained retriever to dynamically retrieve relevant information during text generation. The research team evaluated M+ on a variety of benchmarks, and the experimental results showed that M+ significantly outperformed MemoryLLM and recent strong baseline models, and expanded the knowledge retention capacity from less than 20,000 tokens to more than 160,000 tokens with similar GPU memory overhead.