Memora: A Harmonic Memory Representation Balancing Abstraction and Specificity

Microsoft Research introduces Memora, a scalable agentic memory framework designed to balance abstraction and specificity for long-horizon AI tasks. The system decouples rich memory content from lightweight retrieval structures, setting new state-of-the-art results on benchmarks while using up to 98% fewer context tokens.

• Memora decouples what is stored (rich memory content) from how it is retrieved (lightweight abstractions and cue anchors).
• Each memory entry consists of a primary abstraction (6–8 words) for embedding-based search and a memory value holding the full details.
• Cue anchors provide flexible, context-aware tags as alternative access paths to memories without rigid ontologies.
• The system outperforms Mem0, RAG, and full-context inference on LoCoMo and LongMemEval benchmarks.

This approach resolves the trade-off between preserving fine-grained detail and organizing memory efficiently, allowing agents to navigate their history without re-reading entire conversations.