← Selected WorksGitHub

AI / Research / LLM · 2025

MemoryLens

Open-source LLM memory benchmark — 5.45× recall/token at T=100.

MEMO
// Overview

MemoryLens is an open-source framework for rigorous comparison of LLM memory architectures across long conversations. Three backends — Naive (full context), RAG (semantic retrieval), and Cascading Temporal (hierarchical decay) — are evaluated against five content-based metrics at conversation lengths up to T=100. The headline result: Cascading Temporal achieves 5.45× recall per token over naive at T=100 with 78% lower cost. All results are surfaced in a Streamlit dashboard with automated CI via GitHub Actions.

// Specs

Specifications

TypeOpen-source LLM benchmark
Recall5.45× improvement (Cascading Temporal vs Naive)
Cost Reduction78% at T=100 vs naive
BackendsNaive · RAG · Cascading Temporal
Metrics5 content-based evaluation metrics
// Features

Features

01

3 memory backends: Naive full-context, RAG retrieval, Cascading Temporal

02

5 content-based evaluation metrics for memory quality scoring

03

Cascading Temporal achieves 5.45× recall/token vs naive at T=100

04

78% lower token cost vs naive full-context approach at T=100

05

Streamlit interactive dashboard for result exploration and visualisation

06

CI/CD automated testing and regression checks via GitHub Actions

// Tech

Tech Stack

Groqsentence-transformersStreamlitGitHub Actions

Interested in
MemoryLens?

Explore the source code, architecture notes, and full implementation.

View on GitHubGet in Touch →