Generative modeling with Gaussian Boson Sampling: classically trainable Bosonic Born Machines Quantum generative modeling has emerged as a promising application of quantum computers, aiming to model complex probability distributions beyond the reach of classical methods. In practice, however, ...

Today we close the week with a Gaussian Boson Sampling model that we can train with hundreds of photonic modes and more than a million trainable parameters using a single GPU.
arxiv.org/abs/2603.11195
3/3

Universality of Classically Trainable, Quantum-Deployed Boson-Sampling Generative Models Recent work on the instantaneous quantum polynomial-time (IQP) quantum-circuit Born machine (QCBM) highlights a promising paradigm for generative modeling: train classically, deploy quantumly. In this...

On Wednesday, colleagues from Quandela put out a paper on how to do efficient classical training of Fock-state Boson Sampling. arxiv.org/abs/2603.08793
Yesterday, we put out similar results complementing it with universality guarantees and connections to hardness. arxiv.org/abs/2603.11014 2/3

This week scalable quantum generative learning finally reached bosonic systems:
scirate.com/arxiv/2603.1...
1/3

Wow! I also strongly hope that it stays nonprofit...

Great thread by Stefan Knecht on the new paper by the Garnet Chan group:
"FeMoCo hasn’t been entirely solved, but for its gold-standard active space model (which is the model QC folks have been looking at!), there is now a classical benchmark that provides an excellent baseline for any quantum work"

A larger update of our Majorana propagation paper is out!
arxiv.org/abs/2503.18939

Great! Huge congrats!

With Bence and Zoli, we were able to train Fermionic Born Machine circuits even in the range of 120–160 qubits quite easily, but stay tuned for further improvements to our training methods, where we aim beyond 1000 qubits… 6/6

Our model builds on previous work on the hardness of Fermion Sampling (fermionic magic states + fermionic linear optics) with @michaloszmaniec.bsky.social (Michal Oszmaniec), Mauro Morales, and Ninnat Dangniam 5/6
journals.aps.org/prxquantum/a...

Together with my two fantastic PhD students, Bence Bakó and Zoltán Kolarovszki, we now introduce a second compelling quantum generative model that is also (potentially) hard to sample from classically, yet can be trained classically. 4/6

In that they show that generative models based on IQP circuits can be trained efficiently on classical hardware, while classical sampling remains hard - offering a potential computational advantage and at the same time resolving issues related to quantum gradient estimation. 3/6

Train on classical, deploy on quantum: scaling generative quantum machine learning to a thousand qubits We propose an approach to generative quantum machine learning that overcomes the fundamental scaling issues of variational quantum circuits. The core idea is to use a class of generative models based ...

A related generative QML work that really impressed me this year was by Joseph Bowles, Erik Recio-Armengol, and Shahnawaz Ahmed 2/6
arxiv.org/abs/2503.02934

Happy to see that our paper on the classical training and quantum inference of Fermionic Born Machines is now out on arXiv! 1/6

scirate.com/arxiv/2511.1...

Super happy to see that our paper on the "Quantum generalizations of Glauber and Metropolis dynamics" with Anthony Chen, Csaba Czabán, João Doriguello, András Gilyén, Balázs Kabella, Michael Kastoryano, József Mák was accepted as a talk at #QIP2026 🎉🎉🎉

Thanks a lot for visiting, Alex (@physicistalex.bsky.social)!
It was super nice and productive - we found a couple of nice platonic ideas concerning hyperbolic tilings and spin models.

No-go theorems for sublinear-depth group designs Constructing ensembles of circuits which efficiently approximate the Haar measure over various groups is a long-standing and fundamental problem in quantum information theory. Recently it was shown th...

See See also the great parallel work by M. West, D. García-Martín, N.L. Diaz, @mvscerezo.bsky.social, and @martinlaroo.bsky.social: scirate.com/arxiv/2506.1...

The paper was a result of a wonderful collaboration with Lorenzo Grevink, @haferjonas.bsky.social , @markusheinrich.bsky.social, Jonas Helsen, Marcel Hinsche, Tommy Schuster that started during a discussion at the Random Quantum Circuits in Amsterdam.

In our new paper, we derive linear lower bounds on the circuit depth of Clifford, orthogonal, symplectic, matchgate designs (built of local gates from the respective groups).

This is not the way for HTML papers to gain visibility, but it's a kind of fallback I suppose.

Last day's PDFs should be gradually coming back online, apologies for the delay.

Yes, at least the HTML papers got visibility 😀.
Btw, the pdfs were in deed back in the afternoon. Thx, for reacting fast!

ArXiv seems broken this morning, can't get the pdf of today's papers. (Btw, the html option still works, so one can access the papers in that form.)
@arxiv-quant-ph.bsky.social

Our new Majorana Propagation method is out! It can be used either alone as a classical simulation method or in conjunction with quantum subroutines, see @qzoeholmes.bsky.social's thread below.
(Btw, it's so efficient that it broke scirate: this and all papers after ours are missing from scirate.)

Happy to share that I will join University of Helsinki as a Professor at the Department of Physics. The team at UH has been incredibly welcoming, and I’m looking forward to working here. More to come soon (e.g., announcements of open postdoc, PhD positions)!

Wow! Wonder how what Google map shows in Mexico....

That was fast! Today on my Google Map:

I use my quantum crystals every morning

Recently, I was rereading Carl Sagan’s book, "The Demon-Haunted World: Science as a Candle in the Dark" published in 1995. I am just now randomly posting a paragraph from it; you know, any resemblance to actual events is purely coincidental.

The central idea of our work is that the difficulty of counting degrees of freedom in quantum gravity has a nice quantum info analogue: Verifying the size of a Hilbert space with qubit operations ("overlapping qubits problem") .
Stay tuned for further works on this!

Overlapping qubits from non-isometric maps and de Sitter tensor networks - Nature Communications For some states in quantum gravity, notions of locality can deviate from effective field theory predictions. Here, the authors show that such deviations can be recast in the quantum information framew...

I am super happy that this work is out now in Nature Communications!
It was really fun to work and discuss with Charles ChunJun Cao, Wissam Chemissany and Alexander Jahn!

www.nature.com/articles/s41...

"We do not yet have proven exponential quantum speedups for end-to-end applications in machine learning, optimization, quantum chemistry, or materials science that guarantee substantial commercial and financial value." arxiv.org/pdf/2501.05694