Is LightMem superseded?

LightMem (Agent / long-term memory): superseded — cited as a baseline and beaten by newer methods. 1 paper(s) critique it, 4 beat it on benchmarks — #10 of 63 most-superseded. Sub-problem: cluster led by Mem0. Newer alternatives in the same sub-problem include REAL, MemPro, MemGuard, DeferMem, Memory-R2.

Method Drift›Agent / long-term memory

Superseded baseline#10 of 63 most-superseded

LightMem

LightMem: Lightweight and Efficient Memory-Augmented Generation

Agent / long-term memory · first seen Oct 21, 2025

superseded — cited as a baseline and beaten by newer methods

1 papers critique it · 4 beat it on benchmarks

What papers say

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

“Taking LightMem on LoCoMo as an example: open-domain (75.9% accuracy) and single-hop (70.7% accuracy) questions are handled well, but multi-hop reasoning drops to 60.6% and temporal reasoning to just 45.8%.”
— GRAVITY: Architecture-Agnostic Structured Anchoring for Long-Horizon Conversational Memory

Beaten on benchmarks

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

GRAVITY beats LightMem · LLM-judge accuracy [LongMemEval Micro]
72.6 vs 68.8
GRAVITY: Architecture-Agnostic Structured Anchoring for Long-Horizon Conversational Memory
GRAVITY beats LightMem · LLM-judge accuracy [LongMemEval Macro]
70.9 vs 67.1
GRAVITY: Architecture-Agnostic Structured Anchoring for Long-Horizon Conversational Memory
GRAVITY beats LightMem · LLM-judge accuracy [LoCoMo]
75.8 vs 70.1
GRAVITY: Architecture-Agnostic Structured Anchoring for Long-Horizon Conversational Memory
DeltaMem beats LightMem · LJ [Overall]
75.13 vs 66.43
DeltaMem: Towards Agentic Memory Management via Reinforcement Learning
DeltaMem beats LightMem · QA [Question Answering]
66.43 vs 58.38
DeltaMem: Towards Agentic Memory Management via Reinforcement Learning
GAM beats LightMem · LongBenchv2 F1 [GPT-4o-mini]
73.59 vs 69.51
General Agentic Memory Via Deep Research
DeferMem beats LightMem · Accuracy [LongMemEval-S]
70.00 vs 68.64
DeferMem: Query-Time Evidence Distillation via Reinforcement Learning for Long-Term Memory QA
DeferMem beats LightMem · Token cost [LongMemEval-S]
0.00 vs 28.25
DeferMem: Query-Time Evidence Distillation via Reinforcement Learning for Long-Term Memory QA
DeferMem beats LightMem · Time cost [LongMemEval-S]
90.30 vs 283.76
DeferMem: Query-Time Evidence Distillation via Reinforcement Learning for Long-Term Memory QA
DeferMem beats LightMem · Accuracy [LoCoMo (all convs) non-adversarial]
88.25 vs 72.99
DeferMem: Query-Time Evidence Distillation via Reinforcement Learning for Long-Term Memory QA
DeferMem beats LightMem · Accuracy [LoCoMo (all convs) adversarial]
97.09 vs 90.81
DeferMem: Query-Time Evidence Distillation via Reinforcement Learning for Long-Term Memory QA
DeferMem beats LightMem · Time cost [LoCoMo (all convs) non-adversarial]
83.56 vs 815.32
DeferMem: Query-Time Evidence Distillation via Reinforcement Learning for Long-Term Memory QA

Newer alternatives

Recent methods in the same sub-problem, not yet superseded in the knowledge base.