清华大学与阿里云的研究团队联合开发了一种新的强化学习方法，名为MemPO（Self-Memory Policy Optimization），旨在解决长期交互任务中LLM Agent的上下文膨胀问题。该技术通过引入基于有效信息含量的记忆优势估计机制，使模型能够自主管理和优化其记忆内容，从而提高任务执行效率。

Semiconductor / EDA: One of the World's Leading Chip Manufacturers is accelerating time-to-results across advanced electronic design automation (EDA) workloads. By expanding memory capacity, MEXT ...

Embedded systems demand high performance with minimal power consumption, and the optimisation of scratchpad memory (SPM) plays a critical role in meeting these stringent requirements. SPM, a small ...

Researchers at the Tokyo-based startup Sakana AI have developed a new technique that enables language models to use memory more efficiently, helping enterprises cut the costs of building applications ...

The dynamic interplay between processor speed and memory access times has rendered cache performance a critical determinant of computing efficiency. As modern systems increasingly rely on hierarchical ...

OPENEDGES Technology, a leading provider of memory subsystem IP solutions, today announced that it has secured its first ...

With TurboQuant, Google promises 'massive compression for large language models.' ...

Google's TurboQuant combines PolarQuant with Quantized Johnson-Lindenstrauss correction to shrink memory use, raising ...

Artificial intelligence (AI) has opened up a new can of worms for the tech industry, with memory prices increasing rapidly as demand grows. In response to these increased costs, manufacturers will be ...