An all-transistor physical-learning transformer can be trained to perform in-context learning
Abstract
Most systems trained by equilibrium propagation or coupled learning are in the same class as multi-layer perceptrons: while they may be universal function approximators, they struggle to learn associative tasks which are easy for transformers. Equilibrium propagation relies on a bidirectional flow of information through the physical system between input and output. The reverse flow is broken by more complex nonlinear systems, such as those performing multiplication which is needed to express dot-product attention. Here we explicitly restore this bidirectional flow by introducing extra circuit elements for this purpose into a double or dyadic circuit. We then use these ideas to construct an attention mechanism exploiting the exponential current law of sub-threshold MOSFET devices. We show that in simulations this circuit can be trained to perform a simple in-context learning task, despite using realistic reverse components which transmit only an approximate gradient. For this training to succeed we need cross-entropy loss (not mean squared error). We generate nudges appropriate for this loss using a novel floating circuit element, based on the same MOSFET sub-circuit used for softmax. We train the resistors in this circuit using the Lion optimiser, which can be implemented using one capacitor per parameter to store local memory of past gradient estimates.
Related
- Embodied physical learning: a trained bacterial cell wall can guide organism behavior
- A generative pre-trained transformer with Kerr-soliton attention
- Dual Branch Transformer with Physical Priors for PSF Acquisition to Enable PSF Guided Image Restoration against Co Phasing Errors in Optical Synthetic Aperture Interferometric Imaging
- Learning the Geometry of Materials Space with Trained Neural Tangent Kernels
- Can Machine Learning Potentials Be More Accurate Than the Methods They Are Trained On?
BibTeX
@misc{abbott2026alltransistor,
title = {{An all-transistor physical-learning transformer can be trained to perform in-context learning}},
author = {Michael C. Abbott and Benjamin B. Machta},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.12571},
eprint = {2610.12571},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.12571},
}
RIS
TY - UNPB
TI - An all-transistor physical-learning transformer can be trained to perform in-context learning
AU - Michael C. Abbott
AU - Benjamin B. Machta
PY - 2026
DA - 2026/10/08
PB - arXiv
DO - 10.48550/arxiv.2610.12571
UR - https://arxiv.org/abs/2610.12571
AB - Most systems trained by equilibrium propagation or coupled learning are in the same class as multi-layer perceptrons: while they may be universal function approximators, they struggle to learn associative tasks which are easy for transformers. Equilibrium propagation relies on a bidirectional flow of information through the physical system between input and output. The reverse flow is broken by more complex nonlinear systems, such as those performing multiplication which is needed to express dot-product attention. Here we explicitly restore this bidirectional flow by introducing extra circuit elements for this purpose into a double or dyadic circuit. We then use these ideas to construct an attention mechanism exploiting the exponential current law of sub-threshold MOSFET devices. We show that in simulations this circuit can be trained to perform a simple in-context learning task, despite using realistic reverse components which transmit only an approximate gradient. For this training to succeed we need cross-entropy loss (not mean squared error). We generate nudges appropriate for this loss using a novel floating circuit element, based on the same MOSFET sub-circuit used for softmax. We train the resistors in this circuit using the Lion optimiser, which can be implemented using one capacitor per parameter to store local memory of past gradient estimates.
ER -