Leomrlin commented on issue #773: URL: https://github.com/apache/geaflow/issues/773#issuecomment-4028338418
Hi, @SYaoJun We greatly appreciate your interest in contributing to these vector implementations! Regarding MagnitudeVector: This vector type is designed to capture the LLM's sensitivity to magnitude or quantitative information embedded within memories. For instance, when a user specifies a timeframe like "between 2003 and 2023," this vector could guide the memory system to retrieve content associated with similar or overlapping numerical ranges. Another example would be recalling memories tagged with "temperature above 30°C" when such a condition is mentioned, effectively querying an open interval. Regarding TraversalVector: This vector aims to capture the LLM's understanding of graph semantics to represent graph traversal intents. For example, in a knowledge graph containing entity types like "Person" and "School" connected by a "graduated from" relation, the LLM could formulate a query vector such as (Person_X, graduated from, School). This would instruct the memory system to perform a structured recall based on the graph's connectivity patterns. Currently, this memory system is still in the active research and development phase. We are exploring and refining these concepts, and we warmly welcome you to join our open-source community to collaborate on building it out. Please also note that this is a preliminary draft—these ideas may not yet be fully mature, and we are very open to more compelling suggestions or alternative approaches. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
