> For the complete documentation index, see [llms.txt](https://docs.quantaria.xyz/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.quantaria.xyz/architecture/quantum-path.md).

# Quantum Path

Quantaria models train and run on GPUs. They do **not** run on quantum computers today. This page describes why tensor-network weights give individual layers a path to quantum hardware, and what is and is not done.

{% hint style="warning" %}
**Hardware status.** Quantum layers are tested on circuit simulators and in small hardware experiments. Quantum layers ship in production Quantaria models only where real hardware supports them. Where a result comes from a simulator, Quantaria says so.
{% endhint %}

## Why the format matches

MPS and MPO are the formats physicists use for quantum states and operators. Three facts connect them to circuits:

* A bond of dimension χ carries at most log₂χ bits of entanglement entropy across its cut.
* An MPS with bond dimension 2^k can be prepared by a **staircase circuit** whose gates act on k+1 qubits (Schön et al., [arXiv:quant-ph/0501096](https://arxiv.org/abs/quant-ph/0501096)).
* Because the layer is already in this form, converting it to a circuit does not require re-deriving the model.

## Disentanglers

MERA-style **disentanglers** can move correlations out of an MPO and into parameterized quantum circuits. In published work ([arXiv:2410.17397](https://arxiv.org/abs/2410.17397)), this cut one SmolLM2 attention layer to about 3,000x fewer classical parameters at +0.26% perplexity, with part of inference run on an IBM Heron processor. This is a published result on that paper's model, not a Quantaria result.

## Unitary adapters

Quantaria's Dirac line plans **Cayley-parameterized unitary adapters** at the boundary between circuit and tensor-network sections. The Cayley transform parameterizes a unitary matrix from an unconstrained matrix, so the adapter stays unitary during gradient-based training.

## Dirac layer structure

A Dirac hybrid layer is built as:

1. a quantum circuit,
2. an MPO,
3. a quantum circuit.

## Sequence

| Step | Target                                                                                                                                                 | Status   |
| ---- | ------------------------------------------------------------------------------------------------------------------------------------------------------ | -------- |
| 1    | GPU training and inference of tensor-network models (Hilbert-1)                                                                                        | Live     |
| 2    | Hybrid layers on circuit simulators                                                                                                                    | Research |
| 3    | Selected layers on cloud quantum processors (QPUs)                                                                                                     | Research |
| 4    | Quantum-native training: area-law rank allocation from measured entanglement entropy, DMRG-style sweeps alongside gradient descent, Born-machine heads | Research |

## Born machines

The Born line samples images from an MPS with probability |ψ(x)|², the way a measurement samples a quantum state. This is a research direction and is not part of Hilbert-1.

## Related

* [Tensor-Network Weights](/architecture/tensor-network-weights.md)
* [Research Foundations](/architecture/research-foundations.md)
* [Models](https://github.com/bhavy2908/dataai_docs/tree/main/overview/models.md#dirac)


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation by asking a question.

Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters:

```
GET https://docs.quantaria.xyz/architecture/quantum-path.md?ask=<question>&goal=<user_goal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.