flâneur

Continual Learning in Token Space | Letta

letta.com · 2,456 words · saved by 2 readers

Create, deploy, and manage your agents at scale with Letta Cloud. Build production applications backed by agent microservices with REST APIs. Letta adds memory to your LLM services to give them advanced reasoning capabilities and transparent long-term memory (powered by MemGPT). The biggest gap between AI agents and human intelligence is the ability to learn. Humans continually learn and improve over time, acquire new skills, update their beliefs based on new facts, and modify their behavior to correct for past mistakes. In contrast, most AI agents have an incredible amount of world knowledge, but do not meaningfully get better over time. How do we create AI agents that can continually learn? Traditionally, the concept of “continual learning” for neural networks has been synonymous with weight updates, under the assumption that all learning happens in a connectionist way. The central research questions have focused on catastrophic forgetting (new weight updates causing accidental knowl

BRIEF The continual learning problem in LLM agents is best viewed through the lens of learning in token space: updates to learned context, not weights, should be the primary mechanism for LLM agents to learn from experience. The biggest gap between AI agents and human intelligence is the ability to learn. Humans continually learn and improve over time, acquire new skills, update their beliefs based on new facts, and modify their behavior to correct for past mistakes. In contrast, most AI agents have an incredible amount of world knowledge, but do not meaningfully get better over time. How do w

saved by

related reading