Zep (Agent / long-term memory): heavily superseded — a standard baseline that newer methods routinely beat. 1 paper(s) critique it, 9 beat it on benchmarks — #3 of 63 most-superseded. Sub-problem: cluster led by Mem0. Newer alternatives in the same sub-problem include REAL, MemPro, MemGuard, DeferMem, Memory-R2.

Agent / long-term memory

Zep

Zep: A Temporal Knowledge Graph Architecture for Agent Memory

Heavily superseded#3 of 63 most-superseded · first seen Jan 20, 2025

Heavily superseded — a standard baseline that newer methods routinely beat

1 papers critique it · 9 beat it on benchmarks

What papers say

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

LLMs cannot use memory tools effectively and using such tools increases redundancy in planning and overall tool use.

— MEMTRACK: Evaluating Long-Term Memory and State Tracking in Multi-Platform Dynamic Agent Environments

Beaten on benchmarks

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

HingeMem beats Zep
63.9 vs 16.3
F1 · [Overall]
HingeMem: Boundary Guided Long-Term Memory with Query Adaptive Retrieval for Scalable Dialogues
SwiftMem beats Zep
0.569 vs 0.200
BLEU-1 · [Temporal Reasoning]
SwiftMem: Fast Agentic Memory via Query-aware Indexing
Chronos beats Zep
91.73 vs 57.90
MS · [GPT-4o model, multi-session aggregation]
Chronos: Temporal-Aware Conversational Agents with Structured Event Retrieval for Long-Term Memory
Mem0 beats Zep
0.708 vs 1.292
Latency total p50 (seconds) · [overall]
Mem0: Building Production-Ready AI Agents with Scalable Long-Term Memory
SmartSearch beats Zep
88.4 vs 63.8
Overall · [full benchmark]
SmartSearch: How Ranking Beats Structure for Conversational Memory Retrieval
Mnemosyne beats Zep
54.55 vs 42.80
Overall (%) · [LoCoMo benchmark]
Mnemosyne: An Unsupervised, Human-Inspired Long-Term Memory Architecture for Edge-Based LLMs
GRAVITY beats Zep
60.9 vs 47.8
LLM-judge accuracy · [LongMemEval Macro]
GRAVITY: Architecture-Agnostic Structured Anchoring for Long-Horizon Conversational Memory
DeltaMem beats Zep
75.13 vs 65.99
LJ · [Overall]
DeltaMem: Towards Agentic Memory Management via Reinforcement Learning
gpt-5+NoMem beats Zep
0.206 vs 0.214
Redundancy · [GPT-5]
MEMTRACK: Evaluating Long-Term Memory and State Tracking in Multi-Platform Dynamic Agent Environments
gemini-2.5-pro+NoMem beats Zep
0.144 vs 0.140
Correctness · [Gemini-2.5-pro]
MEMTRACK: Evaluating Long-Term Memory and State Tracking in Multi-Platform Dynamic Agent Environments

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.