A-MEM (Agent / long-term memory): heavily superseded — a standard baseline that newer methods routinely beat. 5 paper(s) critique it, 12 beat it on benchmarks — #2 of 63 most-superseded. Sub-problem: cluster led by Mem0. Newer alternatives in the same sub-problem include REAL, MemPro, MemGuard, DeferMem, Memory-R2.

Is A-MEM superseded? Critiques, benchmarks & alternatives

What papers say

Verbatim critique sentences, each from a paper that cites A-MEM as a baseline.

However, most existing agent-memory approaches still rely on unweighted or weakly weighted relations, where an edge primarily indicates the existence of a connection rather than its query-dependent utility.

— HAGE: Harnessing Agentic Memory via RL-Driven Weighted Graph Evolution

“Current frameworks operate as synchronous, ``append-and-evolve-all'' systems. Every user utterance---regardless of its information density---is forced through the entire memory construction and evolution pipeline. In production, this design inevitably leads to an $O(N^2)$ computational complexity for memory updates as the interaction history grows.”

— D-MEM: Dopamine-Gated Agentic Memory via Reward Prediction Error Routing

“trigger updates based on arbitrary token counts or time-steps rather than semantic completeness, failing to prevent the corruption of stable knowledge by transient dialogue states.”

— GAM: Hierarchical Graph-based Agentic Memory for LLM Agents

“However, they rely on implicit, unstructured associations rather than explicit schemas for modeling information evolution across sessions. This approach can lead to arbitrary links and inconsistent interpretations that are difficult to analyze.”

— Pre-Storage Reasoning for Episodic Memory: Shifting Inference Burden to Memory for Personalized Dialogue

“However, prior work typically organizes memory around associative proximity (e.g., semantic similarity) rather than mechanistic dependency~kiciman2023causal. As a result, such methods can retrieve what occurred but struggle to reason about why, since they lack explicit representations of causal structure”

— MAGMA: A Multi-Graph based Agentic Memory Architecture for AI Agents

Beaten on benchmarks

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

Mem0 beats A-MEM
51.15 vs 18.85
LLM-as-a-Judge score · [Multi-Hop]
Mem0: Building Production-Ready AI Agents with Scalable Long-Term Memory
HAGE beats A-MEM
0.429 vs 0.186
F1 · [Qwen2.5-3B]
HAGE: Harnessing Agentic Memory via RL-Driven Weighted Graph Evolution
AtomMem beats A-MEM
48.7 vs 29.0
Accuracy · [WebWalker]
AtomMem : Learnable Dynamic Agentic Memory with Atomic Memory Operation
D-MEM beats A-MEM
42.7 vs 27.0
Multi Hop F1 · [clean LoCoMo benchmark]
D-MEM: Dopamine-Gated Agentic Memory via Reward Prediction Error Routing
PREMem beats A-MEM
68.03 vs 43.62
LLM-as-a-judge score · [Qwen2.5-14B on LoCoMo]
Pre-Storage Reasoning for Episodic Memory: Shifting Inference Burden to Memory for Personalized Dialogue
MAGMA beats A-MEM
1.47 vs 2.26
Latency (s) · [all methods]
MAGMA: A Multi-Graph based Agentic Memory Architecture for AI Agents
MemGuard beats A-MEM
77.29 vs 59.62
Avg. · [LoCoMo (Base LLM: GPT-4.1-mini, Judge LLM: GPT-4.1)]
MemGuard: Preventing Memory Contamination in Long-Term Memory-Augmented Large Language Models
GAM beats A-MEM
56.62 vs 44.65
LoCoMo Single Hop F1 · [GPT-4o-mini]
General Agentic Memory Via Deep Research
GRAVITY beats A-MEM
63.0 vs 51.9
LLM-judge accuracy · [LongMemEval Macro]
GRAVITY: Architecture-Agnostic Structured Anchoring for Long-Horizon Conversational Memory
AgeMem beats A-MEM
54.31 vs 45.74
Average · [Qwen3-4B-Instruct]
Agentic Memory: Learning Unified Long-Term and Short-Term Memory Management for Large Language Model Agents
REAL beats A-MEM
41.73 vs 35.90
LJ · [DeepSeek-V3]
REAL: A Reasoning-Enhanced Graph Framework for Long-Term Memory Management of LLMs
DeferMem beats A-MEM
70.00 vs 62.60
Accuracy · [LongMemEval-S]
DeferMem: Query-Time Evidence Distillation via Reinforcement Learning for Long-Term Memory QA

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.