Active Memory
A model's ability to carry key user information across conversations without the user manually saving or retrieving it. It involves three layers of mechanisms: memory writing, retrieval, and forgetting.
Timeline (last 90 days)
Active Memory is a capability of Meta Muse Wonder that can carry important information across conversations and recall relevant details when the current task requires them.
Active Memory is a dynamic semantic memory subsystem that runs continuously in the background, continuously tracking user preferences, constraints and historical decisions, and proactively recalling relevant details across different sessions.
Active Memory abstracts historical conversations into structured facts and retrieves and injects only a small number of relevant vectors, with cache reads costing only $0.20 per million tokens, far lower than the cost of repeatedly feeding in historical conversations.
Active Memory refers to a model persisting user preferences, project context, and historical decisions across sessions and actively invoking them in subsequent conversations; Active Focus is a mechanism for maintaining multi-step execution around an established goal, switching strategies after failure, and continuing to push forward when encountering obstacles.
Meta describes two model features: Active Memory carries relevant context across conversations, and Active Focus keeps trying different paths on difficult goals and confirms before critical actions.
Muse Wonder focuses on two major features: Active Memory and Active Focus.
The memory layer is responsible for storing, retrieving, and managing information accumulated by an Agent during interactions, enabling it to maintain continuity across sessions.
All Facts (7)
Meta describes two model features: Active Memory carries relevant context across conversations, and Active Focus keeps trying different paths on difficult goals and confirms before critical actions.
60%UnverifiedActive Memory is a capability of Meta Muse Wonder that can carry important information across conversations and recall relevant details when the current task requires them.
50%UnverifiedActive Memory is a dynamic semantic memory subsystem that runs continuously in the background, continuously tracking user preferences, constraints and historical decisions, and proactively recalling relevant details across different sessions.
50%UnverifiedActive Memory abstracts historical conversations into structured facts and retrieves and injects only a small number of relevant vectors, with cache reads costing only $0.20 per million tokens, far lower than the cost of repeatedly feeding in historical conversations.
50%UnverifiedActive Memory refers to a model persisting user preferences, project context, and historical decisions across sessions and actively invoking them in subsequent conversations; Active Focus is a mechanism for maintaining multi-step execution around an established goal, switching strategies after failure, and continuing to push forward when encountering obstacles.
50%UnverifiedMuse Wonder focuses on two major features: Active Memory and Active Focus.
50%UnverifiedThe memory layer is responsible for storing, retrieving, and managing information accumulated by an Agent during interactions, enabling it to maintain continuity across sessions.
50%