Memp takes inspiration from human cognition to give LLM agents "procedural memory" that can adapt to new tasks and environments.

VentureBeat is a leading source of news, analysis, and insights on technology innovation, startups, and venture capital. Covering topics such as AI, blockchain, gaming, and more, VentureBeat provides  reporting, interviews, and commentary on trends and developments shaping the tech industry. Entrepreneurs, investors, and technology enthusiasts can stay informed about the latest news, funding rounds, and market trends through VentureBeat's coverage.

Venture Beat

Researchers from Zhejiang University and Alibaba Group developed Memp, a framework that gives AI agents "procedural memory" similar to human skill learning. This system allows agents to continuously build, retrieve, and update memories from past experiences, making them more efficient at complex tasks. Unlike other memory frameworks that focus on remembering conversation content, Memp captures reusable "how-to" knowledge across different tasks. The framework addresses the cold-start problem by using evaluation metrics to bootstrap initial memories and demonstrates transferability between large and small models, potentially reducing costs while maintaining performance.

How procedural memory can cut the cost and complexity of AI agents

The case for procedural memory in AI agents