Tiermem banner
FreedomIntelligence FreedomIntelligence

Tiermem

Development community

Description

From Lossy to Verified: A Provenance-Aware Tiered Memory for Agents

Installation

This entry records only its repository, not the path inside it, so there is no exact command to give. Open the source below and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

README

From Lossy to Verified: A Provenance-Aware Tiered Memory for Agents

[![Python 3.10+](https://img.shields.io/badge/python-3.10+-blue.svg)](https://www.python.org/downloads/) [![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT) [![HuggingFace Model](https://img.shields.io/badge/🤗%20HuggingFace-Model-orange)](https://huggingface.co/FreedomIntelligence/TierMem) [![arXiv](https://img.shields.io/badge/arXiv-2602.17913-b31b1b.svg)](https://arxiv.org/abs/2602.17913)

**A memory-augmented LLM system for long-context question answering with intelligent routing between summary-based and raw-retrieval pipelines.**

[Installation](#installation) • [Quick Start](#quick-start) • [Model](#pretrained-model) • [Benchmarks](#supported-benchmarks) • [Training](#router-training)


Table of Contents

Overview

TierMem implements a two-tier memory architecture that balances efficiency and accuracy for long-context question answering:

  1. Summary Index (S-path): Fast semantic search over extracted facts using Mem0
  2. Page Store (R-path): Raw conversation chunks with BM25 retrieval for detailed context

A trained router model dynamically selects between these paths based on query complexity, ensuring optimal performance across different types of questions.

Key Features

✨ **Intelligent Routing** - Trained router automatically selects the best retrieval strategy 🚀 **High Performance** - Optimized fo