Sources and further reading
Primary references for the memory story, representations, and evaluation.
The main pages cite sources beside the relevant ideas. This list supports deeper reading. Research findings apply to the authors' tested settings; documentation describes a particular implementation. Reviewed 18 September 2026.
Memory roles and context
- CoALA, 2023, revised 2024: a framework for language-agent memory and actions.
- LangGraph memory overview, implementation documentation: thread-scoped state and cross-thread memory.
- MemGPT, 2023: managing information across memory levels.
- Generative Agents, 2023: memories, reflections, and planning in a simulated environment, not enterprise validation.
Representations and storage
- Sentence-BERT, 2019: sentence representations.
- CLIP, 2021: aligned image and text representations.
- InternVideo2, 2024: video representation research.
- CLAP, 2022: audio and language representations.
- Google Satellite Embedding dataset documentation, built with Google DeepMind's AlphaEarth Foundations model: annual satellite embeddings and their interpretation.
- TS2Vec, 2021: time-series representations.
- RAPTOR, 2024: tree-organized summaries for retrieval.
- Titans, submitted December 2024: neural memory updated at test time.
Evaluation and scale
- LoCoMo, 2024: conversational memory evaluation.
- LongMemEval, first published 2024: long-term interactive memory abilities.
- BEAM, 2025, revised 2026: generated conversations reaching ten million tokens.
- NoLiMa, 2025: long-context evaluation beyond literal matching.
- MemPalace benchmark notes, project-published: metric and tuning disclosures. These do not independently validate the guide author's reported experiments.
Operational design
- NVIDIA video-search blueprint, vendor documentation.
- Roboflow Workflows, vendor documentation.
- NIST AI RMF Playbook, lifecycle and governance guidance.
Provider prices, current product availability, and benchmark leaderboards require fresh verification for a procurement decision. They are not needed to understand the memory foundations.
Next step: see how memory learnings are promoted under control on the learning flywheel, or return to Start here.
Cite this page
Written by Murali Sid, Founder and AI Systems Architect, last updated 15 October 2026. Licensed CC BY-SA 4.0, so quote it freely with attribution.
Murali Sid (2026). Sources and further reading. Agentic Architecture Skills. https://www.agenticarchitectureskills.com/memory/sourcesMachine-readable version of this page: /memory/sources.md
Research findings and what remains open
Separate author-reported experiments, public evidence, and work still to be reproduced.
Context patterns for agents
Nine tested ways to put the right material in front of an agent. Each says what to do, when it applies, and the evidence behind it. Patterns one to three give most of the value.