Agent / long-term memory
Nemori
Nemori: Self-Organizing Agent Memory Inspired by Cognitive Science
Heavily superseded — a standard baseline that newer methods routinely beat
2 papers critique it · 5 beat it on benchmarks
What papers say
Verbatim critique sentences, each from a paper that cites Nemori as a baseline.
However, in many such systems, memory access still depends on fixed edge types, manually designed weighting rules, or heuristic traversal procedures.
“Causal Flatness: Systems like A-MEM and Nemori organize memory based on associative proximity (e.g., semantic links) rather than mechanistic dependency. They can retrieve what happened but struggle to reason about why, as they lack explicit causal modeling.”
Beaten on benchmarks
Head-to-head results where a newer method reports beating Nemori. Values are copied from the source paper's tables — verify against the cited paper.
HAGE beats Nemori
0.678 vs 0.131
F1 · [GPT-4o-mini]
HAGE: Harnessing Agentic Memory via RL-Driven Weighted Graph EvolutionSwiftMem beats Nemori
11 vs 835
Search latency · [Overall]
SwiftMem: Fast Agentic Memory via Query-aware IndexingMAGMA beats Nemori
1.47 vs 2.59
Latency (s) · [all methods]
MAGMA: A Multi-Graph based Agentic Memory Architecture for AI AgentsSmartSearch beats Nemori
88.4 vs 74.6
Overall · [full benchmark]
SmartSearch: How Ranking Beats Structure for Conversational Memory RetrievalOursGraph beats Nemori
0.853 vs 0.794
Overall LLM Score · [gpt-4.1-mini backbone]
A Simple Yet Strong Baseline for Long-Term Conversational Memory of LLM Agents
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Recent methods in the same sub-problem, not yet superseded in the knowledge base — arXiv benchmark leaders, not vetted production recommendations.
- May 11, 2026
- Mar 16, 2026