CEDL: A hippocampal-inspired language model architecture

The human brain does not treat memory as a flat database. In hippocampal memory circuits, nearby but competing experiences can be separated, stabilized, retrieved, compared, and updated. This matters because many real decisions are not about remembering something once; they are about choosing the current memory when an older memory is still familiar.

Modern language models and agent systems are powerful, but their memory behavior is often still indirect. Transformer attention is excellent at pattern matching over context, while many agent frameworks attach memory through retrieval tools, vector stores, or prompt history. These approaches can work well, but stale facts, updated instructions, and conflicting memories are often handled by prompting or external orchestration rather than by an explicit cognitive-memory pathway.

CEDL explores that missing layer.

CEDL stands for:

• C - Contextual Encoding: builds a context frame for the current sequence.
• E - Expansion: expands and sparsifies representations to support pattern separation.
• D - Densification: retrieves and consolidates memory-like signals into a usable state.
• L - Linkage: compares, links, and routes the final response through memory-aware feedback.

The architecture is inspired by the entorhinal cortex, dentate gyrus, CA3, and CA1 organization of hippocampal processing. It is not a literal brain simulation; it uses neuroscience as an architectural guide for building a testable cognitive-memory model.

This repository provides a research checkpoint for inspecting CEDL, reproducing current-vs-stale memory-update probes, and experimenting with contextual memory readout.

Real-World Example

Imagine you are meeting a colleague in a large conference center. The first message says:

"Meet me in Room A14."

Ten minutes later, a new message arrives:

"The room changed. Meet me in Room C27."

Both memories are now active. Room A14 is familiar because you saw it first. Room C27 is current because it came from the later update. A useful cognitive system must separate the stale memory from the updated one, retrieve both possibilities, and route the current answer.

For CEDL, this is the core mechanism question:

When old and new information conflict, can the model use contextual memory readout to select the current memory rather than the stale one?

The released probes test exactly this kind of current-vs-stale update behavior.

Intended Use

Use this repository to:

• load the CEDL PyTorch checkpoint;
• inspect the CEDL architecture implementation;
• reproduce current-vs-stale memory-update probes;
• study contextual memory readout and memory-off ablations.

Scope

CEDL is best understood as a NeuroAI research artifact: a language model and cognitive-memory framework for studying how an internal memory-readout pathway changes behavior under controlled conflict. The current checkpoint is strongest as mechanistic evidence for contextual memory updating. Its general downstream results are reported transparently below so readers can separate language-model performance from the memory mechanism result.

Keywords

CEDL, NeuroAI, brain-inspired AI, hippocampal-inspired language model, cognitive architecture, contextual memory, memory readout, memory updating, current-vs-stale memory, pattern separation, memory retrieval, hippocampal circuit, entorhinal cortex, dentate gyrus, CA3, CA1, memory-augmented language model, interpretable memory mechanism.

Files

Expected checkpoint files:

• pytorch_model.bin
• cedl_config.json
• MANIFEST.json

Code and probes:

• CEDL.py
• data_v4c_pairs.py
• probes/probe_memory_causality.py
• probes/probe_memory_diagnosis.py
• probes/probe_memory_source_readout.py

Evidence Summary

Current-vs-stale causality:

Seed	N	CEDL active top-1	Memory-off top-1	Delta acc.	Delta margin	Cohen's d
1	158	79.1%	0.0%	+79.1pp	+8.266	1.656
2	164	73.2%	0.0%	+73.2pp	+7.702	1.481
3	162	72.8%	0.0%	+72.8pp	+7.756	1.459

Memory-readout diagnosis:

Seed	N	Trunk current top-1	Memory current top-1	Memory stale top-1	Current median rank
1	158	0.0%	79.7%	0.6%	1
2	164	0.0%	73.8%	0.6%	1
3	162	0.0%	73.5%	1.2%	1

General evaluation:

Model	Params	WikiText-103 PPL	LAMBADA	HellaSwag	ARC-Easy
CEDL	136.3M	30.11	15.9%	27.1%	28.1%
Transformer baseline	103.2M	45.93	3.9%	26.5%	28.2%

About the Author

CEDL was developed by Dian Jiao, a multidisciplinary researcher and technology leader with more than 15 years of experience across digital transformation, neuroscience, and technology innovation. His work connects brain-inspired theory with practical AI systems, with a focus on NeuroAI, digital therapeutics, and human-centered model design.

His research interests include AI-enhanced digital therapeutics using EEG, adaptive biofeedback, and machine learning for cognitive enhancement and neurological rehabilitation; and biologically inspired AI models that draw from neural dynamics, memory circuits, and brain mechanisms.

Selected publications and contributions:

• Jiao, D. (2025). Elliptic cortical networks: A mathematically constrained architecture for biologically-inspired intelligence. Neurocomputing, 658, 131802. doi: 10.1016/j.neucom.2025.131802
• Jiao, D. (2025). Leveraging neurotechnology for neurodivergent education: a narrative review. Learning: Research and Practice, 1-25. doi: 10.1080/23735082.2025.2517052
• Jiao, D. (2025). AI-enhanced digital therapeutics for cognitive impairment: Integrating mobile applications, virtual reality, and wearable devices. Discover Artificial Intelligence, 5, Article 69. doi: 10.1007/s44163-025-00325-6
• Jiao, D. (2025). Advancing personalized digital therapeutics: integrating music therapy, brainwave entrainment methods, and AI-driven biofeedback. Frontiers in Digital Health, 7, 1552396. doi: 10.3389/fdgth.2025.1552396
• Jiao, D. (2025). From hypoxic pockets to daily routines: linking brain oxygenation and cognitive resilience. Frontiers in Aging Neuroscience, 17, 1534198. doi: 10.3389/fnagi.2025.1534198
• Cognition and Brain Teaching Technology, a book on brain-based teaching and cognitive technology.
• Chinese translations of How Breakthroughs Happen (2020), Mastering the Dynamics of Innovation (2022), and Neuroscience of You (2025).

LinkedIn: https://www.linkedin.com/in/jason-jiao-2972141a4

Load Example

pip install -r requirements.txt
python examples/load_checkpoint.py --checkpoint pytorch_model.bin

Probe Example

python probes/probe_memory_causality.py \
  --checkpoint pytorch_model.bin \
  --sidecar cedl_config.json \
  --n-items 200 \
  --seed 1

Citation

If you use this checkpoint before the manuscript is published, cite it as:

@misc{jiao2026cedl,
  title = {CEDL: A Hippocampal-Inspired Language Model with Contextual Memory Readout},
  author = {Jiao, Dian},
  year = {2026},
  note = {Research checkpoint for contextual memory-readout probes}
}