Manuel Rudolph

Welcoming summer the best way we know how: with pasta, physics, and a phenomenal team 🍝⚛️

A warm #Google #Quantum #AI welcome to Manuel Rudolph, who’s joining us this summer! 🎉 We’re thrilled to have his sharp mind and curious spirit with us

Thx Nikita+team for organizing

Huge thanks to @joeytindall.bsky.social. I had a great time working with him during my 4-month stay at the Flatiron Institute in New York. This guy is amazing!

A big thank you also to @mstoud.bsky.social for having me.

@flatironinstitute.org @simonsfoundation.org

Particularly interesting to us was witnessing how slowly loop correlations build up in heavy-hex processors. Loop correlations are what make loopy networks potentially significantly harder to use than loop-free MPS and tree-tensor networks.

We also introduce a bunch of metrics to certify that the samples are of high-quality. This way, we verified that we solved the biggest circuit in IBM's recent quantum chemistry experiment to numerical precision.

We combine some existing ideas with ITensorNetworks.jl and @joeytindall.bsky.social's flexible boundary MPS code to adapt to any planar geometry.

With our open-source (but not completely polished) software, you can start simulating and sampling 2D circuits today: github.com/JoeyT1994/Te...

In case you thought you can't efficiently simulate and sample quantum circuits with 2D tensor networks... Nope, you can.

Link: scirate.com/arxiv/2507.1...

We simulate IBM's recent quantum chemistry experiments and Williow + heavy-hex Heisenberg dynamics, and showcase modern, verifiable techniques.

Big congrats to @carrasqu.bsky.social's group at @ethz.ch, including Yuxuan Zhang and Roeland Wiersema, and particularly Matteo D'Anna for his amazing first-author work early in his PhD.

Oh no, haha. My bad and thanks!

I don't see any reason why not every circuit executed on hardware should be compressed. The approach can also be used to re-compile into a different gate set or topology.

Classical simulation is not just here to compete with quantum devices.

Last week, we published a paper that really excites me:
"Circuit compression for 2D quantum dynamics"

Using Pauli propagation, we (Matteo) were able to compress Trotter circuits for systems up to 30x30 with depth reductions of x2 to x13.

Link: arxiv.org/abs/2507.01883

I'm surprised Grover's is on here. Maybe not for database search but as a form of amplitude amplification?

Retweeting this for the folks (myself included) that weren't online over the long weekend

PauliPropagation.jl is open source library that you can use to approximately simulate quantum circuits

We explain the nitty gritty of how these algorithms work in practise in our latest companion paper

🧵👇

Sorry 🤐

This paper was meant to go live yesterday (still love you arXiv), but who doesn't scroll social media on a holiday 💁

Thanks to my amazing group and co-authors Tyson Jones, Yanting Teng (@yteng.bsky.social), Armando Angrisani (@aangrisani.bsky.social), and Zoë Holmes (@qzoeholmes.bsky.social)!

Pauli propagation is naturally interfaced with both quantum computers and other classical simulation methods - the perfect team player!

I love improving classical algorithms for simulating quantum computations, and I truly believe performant classical methods are good for everyone.

In VERY short:

- PP is a recent path integral method that is orthogonal to e.g. tensor networks.
- PP evolves objects that are sparse Pauli basis, commonly observables in the Heisenberg picture.
- PP is amazing for quick estimates in low-ish Magic quantum systems.
- PP is hard to converge exactly.

This "compact" 30-page main text manuscript summarizes much of what we have learned about Pauli propagation (PP) and where its strengths lie.

From the general framework description, over theoretical guarantees, to the nitty-gritty implementation details that you are happy to not have to deal with.

❗New paper and open-source library❗

PauliPropagation.jl is your go-to library for simulating quantum circuits via Pauli propagation. Our paper provides a thorough overview of this new classical simulation method.

Paper: scirate.com/arxiv/2505.21606
Library: github.com/MSRudolph/PauliPropagation.jl

Here the general link to UnitaryHACK: unitaryhack.dev

Here the link to our project page: unitaryhack.dev/projects/pau...

And keep your eyes peeled for tomorrow 👁️👁️

UnitaryHACK 2025 has begun - and we are part of it!

If you are registered, earn real money by closing GitHub issues in our new library PauliPropagation.jl (github.com/MSRudolph/PauliPropagation.jl).

We were supposed to have a nice and "compact" paper out today, but the arXiv gods were not with us.

I'm heading to APS for the week. If you are around, let me know!

Or say Hi after my talk in the C34 session on Monday afternoon.

And precisely at that time with a specific approach.

What do you mean by "spoofing"? When I say that I mean that something was a cheap or fake result that looks real. However, at least the TN results are from actual converging simulations (I can't confidently speak of the other EPFL one (shame on me)).

Initially I thought that @joeytindall.bsky.social's paper should have a spicier title, but now I understand that it was never the main point to debunk DWave. They present state-of-the-art circuit simulation techniques in 2D and 3D, and the benchmarks are from one of the best experimental papers.

It seems like proof by exhaustion is a real thing.

Oh no, tensor networks have out-supremed D-Wave!

Dynamics of disordered quantum systems with two- and three-dimensional tensor networks Quantum spin glasses form a good testbed for studying the performance of various quantum annealing and optimization algorithms. In this work we show how two- and three-dimensional tensor networks can ...

In a new preprint arxiv.org/abs/2503.05693, led by Joseph Tindall and Antonio Mello at Flatiron CCQ, we simulate annealing of disordered quantum magnets 🧲 ⌛ and in many cases find better accuracy than recent results from D-Wave devices and leading classical methods (c.f. arxiv.org/abs/2403.00910).

Here we take steps to understanding the potential of warm starts for VQAs

We provide a general variance lower bound for patches of loss landscapes:
- for both structured and unstructured circuits
- to provide small-angle-initialization 'guarantees'
- to study the scaling of regions of attraction

New paper on arXiv 🔥
We present a bound that unifies all the previous guarantees of small regions with substantial gradients in BP landscapes. This allows us to study new architectures, parameter correlations, and points on the landscape that could not be analyzed before.
scirate.com/arxiv/2502.0...

No way we visit IBM's Yorktown office and don't pose for a silly picture in front of the quantum computer.