MiniMem is a deliberately small, transparent baseline for conversational memory evaluation on LoCoMo-style question answering, using LLM fact extraction, embedding cosine similarity, direct top-k retrieval, and token-level F1.

Stars

5

7-day growth

No data

Forks

0

Open issues

0

License

MIT

Last updated

2026-07-30

AI repository intelligence
FR-AI / ANALYSIS

Why it is worth attention

It provides a minimal, understandable reference point in a field dominated by complex memory systems, and includes a curated survey of open-source memory research and a ready-to-run LoCoMo benchmark.

Who it is for

  • Researchers evaluating conversational memory systems
  • Developers seeking a simple baseline for LoCoMo-style QA
  • Educators teaching LLM memory and retrieval pipelines
  • Practitioners comparing lightweight memory pipelines before adopting richer systems

Use cases

  • Baseline scoring on the bundled LoCoMo dataset
  • Debugging and validating LLM fact extraction and retrieval behavior
  • Educational demonstrations of a complete memory QA pipeline
  • Lightweight evaluation of OpenAI-compatible chat models on memory tasks

Strengths

  • Deliberately small pipeline with clearly separated construction, retrieval, answering, and evaluation stages
  • Includes a bundled LoCoMo dataset and scripts to run full evaluation
  • Quickstart makes only two LLM calls and prints progress before each API stage
  • Open-source research index in references/README.md provides context and scope limitations

Considerations

  • Excludes memory updating, clustering, keyword search, reranking, graph retrieval, and token-budget optimization
  • Full LoCoMo evaluation requires 272 memory-construction calls and 1,540 answer calls, so API costs and rate limits must be considered
  • Uses token-level F1 only and evaluates non-adversarial questions, not adversarial ones

README quick start

Quick start

Related repositories

Similar projects matched by category, topics, and programming language.

makecindy
Featured
makecindy GitHub avatar

cindy

Cindy is an open-source AI agent that runs locally on your machine, integrates multiple AI harnesses and models, and provides memory, skills, and automation to perform real work in your projects and apps.

AI & Machine LearningLarge Language Models
958
uzairansaruzi
Featured
uzairansaruzi GitHub avatar

hermex

Hermex is a native SwiftUI iPhone app that lets you control a self-hosted Hermes AI agent directly from your phone, with no subscriptions, tracking, or third-party relay.

AI & Machine LearningLarge Language Models
941
hahhforest
Featured
hahhforest GitHub avatar

pi-textbook

A hands-on course that walks through building a Pi-style coding agent from scratch across 15 checkpoints, starting from an offline agent trajectory.

AI & Machine LearningLarge Language Models
625