Scholar Feed

Agent / long-term memory

MemoryOS

Memory OS of AI Agent

Superseded baseline#8 of 63 most-superseded · first seen May 30, 2025

Superseded — cited as a baseline and beaten by newer methods

2 papers critique it · 4 beat it on benchmarks

What papers say

Verbatim critique sentences, each from a paper that cites MemoryOS as a baseline.

By treating outgoing connections as equally valid or using fixed graph-expansion rules, existing systems can fail to discriminate between highly relevant pathways and distracting noise, leading to degraded retrieval accuracy as memory grows.
— HAGE: Harnessing Agentic Memory via RL-Driven Weighted Graph Evolution
“Blocking Latency: Achieving structural depth often comes at the cost of interactivity. Approaches like MemoryOS and synchronous graph builders typically require heavy LLM operations on the critical path. As noted in benchmarks wu2024longmemeval, such mechanisms incur prohibitive latency, rendering them impractical for real-time interaction.”
— MAGMA: A Multi-Graph based Agentic Memory Architecture for AI Agents

Beaten on benchmarks

Head-to-head results where a newer method reports beating MemoryOS. Values are copied from the source paper's tables — verify against the cited paper.

What to use instead

Recent methods in the same sub-problem, not yet superseded in the knowledge base — arXiv benchmark leaders, not vetted production recommendations.