Reading Template: Memory-augmented Agents
A reusable structure for reading papers about memory, retrieval, and long-horizon agents.
Paper: Replace with the real paper title
Authors: Replace with author names
Venue: Paper Note
Link: Open paper
- Good memory is selective, not just large.
- Retrieval quality matters more than storing everything.
- Long-horizon agents need evaluation over trajectories, not only final answers.
One-sentence Summary
This paper studies how agents can use memory or retrieval to maintain useful context over long interactions.
Problem
What problem does the paper solve?
Write it in your own words. Avoid copying the abstract.
Method
Describe the core method in three to five bullet points.
What I Think Is Important
Focus on what this paper changes in your understanding of AI systems.
Connection to My Work
How can this idea help with multilingual agents, RAG systems, evaluation, or long-horizon task completion?
Critique
What is missing? What assumption may fail in real production systems?