# Graph Memory Transformer (GMT)

arXiv: https://arxiv.org/abs/2604.23862 · HTML: https://mlsanity.com/paper/2604.23862/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 11 of its 46 references that are in the map.
- First posted on arXiv 26 Apr 2026.

## Builds on (most cited first, 11 of 11)

- [Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challenge](https://mlsanity.com/paper/1803.05457/) (2018)
- [Sparse Autoencoders Find Highly Interpretable Features in Language Models](https://mlsanity.com/paper/2309.08600/) (2023)
- [Interpretability in the Wild: a Circuit for Indirect Object Identification in GPT-2 small](https://mlsanity.com/paper/2211.00593/) (2022)
- [Neural Turing Machines](https://mlsanity.com/paper/1410.5401/) (2014)
- [A Comprehensive Study of Knowledge Editing for Large Language Models](https://mlsanity.com/paper/2401.01286/) (2024)
- [A Practical Review of Mechanistic Interpretability for Transformer-Based Language Models](https://mlsanity.com/paper/2407.02646/) (2024)
- [MemoryFormer: Minimize Transformer Computation by Removing Fully-Connected Layers](https://mlsanity.com/paper/2411.12992/) (2024)
- [Neuron-Guided Interpretation of Code LLMs: Where, Why, and How?](https://mlsanity.com/paper/2512.19980/) (2025)
- [MemoryLLM: Plug-n-Play Interpretable Feed-Forward Memory for Transformers](https://mlsanity.com/paper/2602.00398/) (2026)
- [Exploring Activation Patterns of Parameters in Language Models](https://mlsanity.com/paper/2405.17799/) (2024)
- [Prototype Transformer: Towards Language Model Architectures Interpretable by Design](https://mlsanity.com/paper/2602.11852/) (2026)

## Abstract (arXiv)

We investigate whether the Feed-Forward Network (FFN) sublayer in a decoder-only transformer can be replaced by an explicit learned memory graph while preserving the surrounding autoregressive architecture. The proposed Graph Memory Transformer (GMT) keeps causal self-attention intact, but replaces the usual per-token FFN transformation with a memory cell that routes token representations over a learned bank of centroids connected by a learned directed transition matrix. In the base GMT v7 instantiation studied here, each of 16 transformer blocks contains 128 centroids, a 128 * 128 edge matrix, gravitational source routing, token-conditioned target selection, and a gated displacement readout. The cell therefore returns movement from an estimated source memory state toward a target memory state, rather than a retrieved value. The resulting model is a fully decoder-only language model with 82.2M trainable parameters and no dense FFN sublayers, compared with a 103.0M-parameter dense GPT-style baseline used in the evaluation. The base v7 model trains stably and exposes centroid usage, transition structure, and source-to-target movement as directly inspectable quantities of the forward computation. It remains behind the larger dense baseline in validation loss and perplexity (3.5995/36.58 vs. 3.2903/26.85), while showing close zero-shot benchmark behavior under the evaluated setting. These results are not intended as a state-of-the-art claim; they support the viability and structural interpretability of replacing dense within-token transformation with graph-mediated memory navigation. Broader scaling, optimized kernels, and more extensive benchmark evaluation are left for subsequent work.