R&D advanced 3 min read Jun 30, 2026
Public Preview Sign in free for the full digest →

No Agent Memory Architecture Wins Everywhere: 12 Benchmarked

“Zep and Cognee score highest on agent memory — at 155 and 116 seconds per query. LightMem scores 4x lower and runs in 3.67 seconds. Now you have the numbers to make the call.”

No Agent Memory Architecture Wins Everywhere: 12 Benchmarked
Source · alphaxiv.org

“"no single architecture dominates across all scenarios; instead, effectiveness depends heavily on how well the memory structure aligns with the workload bottleneck." — Wei Zhou et al., arXiv:2606.24775, June 2026”

You know that feeling when you read benchmark results from Mem0, Zep, and Cognee and all three claim to be the fastest or most accurate agent memory solution? Each publishes numbers on different tasks, making apples-to-apples comparison impossible. Prior evaluations treat memory as a black box and measure only whether an agent finishes a downstream task — not whether the memory component itself is retrieving, updating, and maintaining facts correctly. The result: you pick a memory architecture based on vendor documentation, deploy it, and discover it fails on the specific workload your agent actually runs, with no diagnostic framework to explain why.

agent-memoryllmbenchmarkresearch-paperpythonopen-sourceretrieval

The paper decomposes every agent memory system into four modules: how it stores information (representation), how it extracts memories from conversation (extraction), how it finds relevant memories at query time (retrieval and routing), and how it keeps memories current (maintenance). For each of the 12 systems, the researchers run ablations that swap one module at a time while keeping the others fixed — so when LightMem's extraction changes from raw to summarized, you see exactly how much accuracy that step destroys (24.2 EM drops to 8.5). Five workload families cover distinct scenarios: accurate retrieval, conflict resolution, test-time learning, long-horizon dialogue, and stateful procedural execution. The benchmark reports three independent axes per system: effectiveness (EM, F1, ROUGE-L), operational cost in seconds per query, and robustness after knowledge updates. That multi-axis view surfaces the trade-off hidden inside single-number headlines.

01
Four-module ablation framework — you can isolate exactly which module (storage, extraction, retrieval, or maintenance) causes your memory system to fail, instead of guessing at a black box when accuracy drops.
02
Cost-utility frontier data — Figure 11 plots query latency versus accuracy for all 12 systems, giving you a principled basis to choose between LightMem (3.67s, 48.3 utility) and Zep (155.1s, 84+ utility) based on your actual latency budget.
03
22 method presets in one benchmark runner — a single command evaluates any of 22 memory architectures including Mem0, Cognee, Letta, GraphRAG, and HippoRAG without writing integration code for each.
04
Temporal update robustness scores — the paper measures what happens after you update a fact in memory, catching systems that return stale answers after knowledge changes (Zep leads at 44.4 EM on update robustness).
05
Late filtering design principle — the finding that broad write-time extraction beats aggressive early filtering gives you a concrete architectural heuristic when building or evaluating any custom memory layer.
06
Five workload families with distinct benchmarks — each maps to a real production scenario (accurate retrieval, conflict resolution, test-time learning, long-horizon dialogue, stateful execution) so you can pick the evaluation that matches ...
Who it’s for

If you are building an agent that needs persistent memory across sessions — a customer support bot, a coding assistant, a personal AI — this paper gives you a framework to evaluate which memory architecture fits your workload before committing. It is also essential reading if you maintain or are evaluating Mem0, Zep, Cognee, or Letta in production and want to understand where each breaks down. Not useful if you need real-time agentic benchmarks (all evaluation is offline and API-dependent) or if your agent has no persistent memory requirement.

Worth exploring

Read this immediately if you are choosing between agent memory architectures — the benchmark surfaces accuracy-latency trade-offs that vendor documentation omits. The code (Python, 72 stars, 0 open issues) is functional but 7 days old with no community reproduction reports yet; treat it as a research resource, not a production-hardened evaluation suite. Revisit in Q3 2026 once the community has attempted reproduction and filed issues.

Developer playbook
Tech stack, code snippet, sentiment, alternatives.
PM playbook
Adoption angles, user fit, positioning.
CEO playbook
Traction signals, ROI, build vs buy.
Deep-dive insight
Full long-form analysis, no fluff.
Easy mode
Core idea, fast — when you need the gist.
Pro mode
Technical nuance, edge cases, tradeoffs.
Read the full digest
Go beyond the preview

Deep-dive insight, Easy and Pro modes, plus action playbooks — the full breakdown is one tap away.

Underrated tools. Unfiltered takes.

Read the full digest in the Snaplyze app for deep-dive insight, Easy and Pro modes, and the playbooks you can actually use.

Install Snaplyze →