Marco Cerezo @mvscerezo.bsky.social · 29/09/2026Its been explicitly told to double check it's math, but maybe it's slacking. I will try someone else's account tomorrow and explicitly test this! [Mine already probably is aware of this error] 010
Reposted by Marco CerezoMarco Cerezo @mvscerezo.bsky.social · 29/09/2026Yeah! For reference, the lesson here is very clear, the bug appears because GPT is extrapolating results from "standard" (holonomic) rep theory to "non standard" (nonholonomic) one. So you can see how the training data REALLY biases the model!!!! Its quite fascinating and a lesson go keep in mind 221
Marco Cerezo @mvscerezo.bsky.social · 29/09/2026@markwilde.bsky.social I haven't seen errors on other stuff tho, and I am reproducing all line by line (and learning a ton along the way hahahahaha!) 010
Marco Cerezo @mvscerezo.bsky.social · 29/09/2026@markwilde.bsky.social I think that as long as you stay within standard stuff you are good, but the edges are very rough once you get to them. The training really pulls the model on wrong directions 000
Marco Cerezo @mvscerezo.bsky.social · 29/09/2026Yeah! For reference, the lesson here is very clear, the bug appears because GPT is extrapolating results from "standard" (holonomic) rep theory to "non standard" (nonholonomic) one. So you can see how the training data REALLY biases the model!!!! Its quite fascinating and a lesson go keep in mind 221
Marco Cerezo @mvscerezo.bsky.social · 29/09/2026Worst part, it wasn't some fancy math. Just basic rep theory of GL(d) on quantum mechanics. I.e. linear algebra. Once I realized the error I tried asking it to justify the wrong math and it keep pushing very confidently until I pointed out the error. Scary! 140
Marco Cerezo @mvscerezo.bsky.social · 29/09/2026⚠️⚠️And this us why we still verify every single proof line by gpt: Today I found a bug 🐛 in a theoretical proof by 5.6 Atra, highest reasoning level! 🤖 It bleeds!! I still matter 😅😅 160
Marco Cerezo @mvscerezo.bsky.social · 25/07/2026We know LLMs can one-shot quantim information papers now, and even write decent manuscripts. What's the consensus of proper practice for what to do with those AI-generated papers? I review the proof, I sign my name, acknowledge the it is AI made, put it on arxiv, rinse, repeat? 2103
Reposted by Marco CerezoNahuel L Diaz @nahuelldiaz.bsky.social · 13/06/2026I'm so happy to share this paper, we put a lot into this one: arxiv.org/abs/2606.12539 If you have ever wondered what the quantum-information literature has to say about temporal correlations and time itself, check out our unifying picture: all you need are spacetime states.arxiv.orgUnifying spacetime approaches to quantum mechanicsRecent efforts to formulate quantum mechanics in a way that treats space and time on a more equal footing have led to a large variety of spacetime-oriented approaches. In this work we present a detail... 274
Marco Cerezo @mvscerezo.bsky.social · 24/05/2026I don’t topically post about my personal life but here it goes. A little personal news: Our baby Emma Lily Cerezo is here ❤️ Healthy happy and already keeping us on our toes. Feeling incredibly grateful and very excited for this new adventure. 6300
Marco Cerezo @mvscerezo.bsky.social · 22/05/2026Please consider submitting your work for this workshop in IEEE Quantum Week 2026 020
Marco Cerezo @mvscerezo.bsky.social · 06/05/202610/ This work started during the Los Alamos Summer School 2024. Huge thanks to my collaborators Paolo Braccia, Pablo Bermejo, @manuqphys.bsky.social , @dgarciamartin.bsky.social , and Jonas Jager!! 030
Marco Cerezo @mvscerezo.bsky.social · 06/05/20269/ We also apply QGPs to a realistic quantum sensing task using sample-efficient Bayesian optimization, where they consistently outperform Hamiltonian learning. 120
Marco Cerezo @mvscerezo.bsky.social · 06/05/20268/ On the applications side, QGPs accurately predict phase diagrams in many-body systems up to 100 qubits. They also perform well for regression, especially extrapolation, classification, and optimization. 130
Marco Cerezo @mvscerezo.bsky.social · 06/05/20267/ On the theory side, we go beyond standard central limit arguments. We also identify natural regimes where the learned process provably deviates from Gaussian behavior, even asymptotically. That is, not everything is a QGP! 110
Marco Cerezo @mvscerezo.bsky.social · 06/05/20266/ Why matchgates? They are free-fermionic transformations that can act non-trivially on all qubits while remaining scalable. This gives QGPs a new kernel that is both physically meaningful and computationally practical. 110
Marco Cerezo @mvscerezo.bsky.social · 06/05/20265/ Next, pick a kernel from the QGP toolbox. Existing options include few-qubit unitary, orthogonal, and symplectic kernels. In this work, we introduce an all-qubit matchgate kernel. 110
Marco Cerezo @mvscerezo.bsky.social · 06/05/20264/ Think of the setup as a recipe. First, gather prior information about the unknown transformation and measurement operators. Then add your quantum dataset, input states plus the measurements observed after the transformation. 120
Marco Cerezo @mvscerezo.bsky.social · 06/05/20263/ QGPs let you encode prior knowledge about a quantum system directly into the model. The result is a Bayesian surrogate that can learn from quantum inputs and measurement outcomes while carrying strong physics-informed inductive biases. 130
Marco Cerezo @mvscerezo.bsky.social · 06/05/20261/ Variational Quantum Machine Learning has hit real roadblocks, from hard optimization to dimensionality curses in parameterized circuits. We thus need quantum learning frameworks that are interpretable, scalable, and aligned with physics, not just more variational stuff! 110
Marco Cerezo @mvscerezo.bsky.social · 06/05/2026⚠️New paper out from our Quantum Computing Summer School! We introduce Quantum Gaussian Processes, QGPs, a Bayesian framework for learning directly from quantum data. arxiv.org/abs/2605.00099 see 🧵 1122
Marco Cerezo @mvscerezo.bsky.social · 06/04/2026Is it me, or is LinkedIn a much better Quantum hub than Twitter or Bluesky? 260
Marco Cerezo @mvscerezo.bsky.social · 06/04/2026Huge thanks to my colleagues Pablo Bermejo, Paolo Braccia, @quantummanuel.bsky.social , @qzoeholmes.bsky.social and Lukasz Cincio 030
Marco Cerezo @mvscerezo.bsky.social · 06/04/2026Our main two messages: 1) Quantum CNNs are VERY easy to simulate. I don't think we can achieve any meaningful quantum advantage from them. Prove me wrong! 2) Our community is in DIRE need of datasets, almost all quantum and classical datasets we use are trivial to solve. 140
Marco Cerezo @mvscerezo.bsky.social · 06/04/2026Extremely proud to see our paper "Quantum Convolutional Neural Networks are Effectively Classically Simulable" published in PRX Quantum journals.aps.org/prxquantum/a... This is an instantiation of our work, provable absence of barren plateaus implies classical simulability.journals.aps.org 2143
Reposted by Marco CerezoDiego García-Martín @dgarciamartin.bsky.social · 25/03/2026New paper out!! 🥳😄 Last Friday we uploaded our latest preprint to the arXiv: "The commutant of fermionic Gaussian unitaries" Amazing collaboration with @impolster.bsky.social, @nahuelldiaz.bsky.social, @martinlaroo.bsky.social and @mvscerezo.bsky.social ! Thread 👇 arxiv.org/abs/2603.19210arxiv.orgThe commutant of fermionic Gaussian unitariesIn this work, we characterize the $t$-th order commutants of fermionic Gaussian unitaries and of their particle-preserving subgroup acting on $n$ fermionic modes. These commutants govern Haar averages... 2133
Marco Cerezo @mvscerezo.bsky.social · 06/03/2026PSA for my friends in quantum 👋 New Mexico Quantum Technologies Award is open—rolling apps thru April. Up to $200K for NM-based companies or folks expanding into NM. Apply: NM Quantum Technologies Award – New Mexico EDD edd.newmexico.gov/quantum-gran...edd.newmexico.govNM Quantum Technologies Award - New Mexico EDDThe New Mexico Quantum Technologies Award (NMQTA) is a competitive, state-funded grant that awards non-dilutive funds to early-stage companies working in 051
Marco Cerezo @mvscerezo.bsky.social · 26/02/2026This will be an awesome conference in an awesome place!! Please consider submitting an abstract! 010
Reposted by Marco Cerezobertacasas.bsky.social @bertacasas.bsky.social · 12/02/2026New preprint: “Matchgate synthesis via Clifford matchgates and T gates.” We propose a new framework for matchgate synthesis: compiling these unitaries using only matchgates. arxiv.org/abs/2602.05425 171
Marco Cerezo @mvscerezo.bsky.social · 04/02/2026I am extremely excited to share our latest news: LANL formed a Center for Quantum Computing!! This Center will enhance our quantum workforce via co-location, and will even host the Summer School! So, if you applied for it, you might be the first cohort in the Center! www.lanl.gov/media/news/0...lanl.govLos Alamos forms quantum computing-focused research center | LANLLos Alamos National Lab forms quantum computing center, consolidating diverse research to drive innovation in national security, algorithms, and workforce development. 1254
Marco Cerezo @mvscerezo.bsky.social · 03/02/2026This method can be combined with ANY local compiler, so any improvement therein, becomes an improvement for us too! 030
Marco Cerezo @mvscerezo.bsky.social · 03/02/2026We use classical AI [Reinforcement Learning] to help quantum circuits: learn local rewrites before synthesis so compilers spit out lower T-count (T gates are $$$ in FTQC). Up to ~20% T reductions, no extra approx error. 120
Marco Cerezo @mvscerezo.bsky.social · 03/02/2026New paper out! 🎆 In collaboration with University of Udine 🇮🇹🤝 arxiv.org/abs/2601.19738 Huge props to my colleagues Daniele Lizzio Bosco, Lukasz Cincio and Giuseppe Serraarxiv.orgQuantum Circuit Pre-Synthesis: Learning Local Edits to Reduce $T$-countCompiling quantum circuits into Clifford+$T$ gates is a central task for fault-tolerant quantum computing using stabilizer codes. In the near term, $T$ gates will dominate the cost of fault tolerant i... 161
Marco Cerezo @mvscerezo.bsky.social · 28/01/2026We still have open postdoc positions in our group! Looking for someone with expertise in quantum computing, tensor networks, Monte Carlo and other methods to simulate quantum many-body systems! Shares appreciates! Apply here: lanl.jobs/search/jobde...lanl.jobsQuantum Computing and Early Fault-Tolerant Simulations at Los Alamos National LaboratoryLos Alamos National Laboratory is Hiring! Search available jobs or submit your resume now by visiting this link. Please share with anyone you feel would be a great fit. 175
Reposted by Marco CerezoMark M. Wilde @markwilde.bsky.social · 15/01/2026Joint work "Genuine multipartite Rains entanglement" w/ Hailey Murray, Sagnik Bhattacharya, M. Cerezo @mvscerezo.bsky.social, Liuke Lyu now available on arXiv: arxiv.org/abs/2601.09590 We establish a genuine multipartite generalization of the famous Rains entanglement measure for bipartite statesarxiv.orgGenuine multipartite Rains entanglementWe introduce the genuine multipartite Rains entanglement (GMRE) as a measure of genuine multipartite entanglement that can be computed using semi-definite programming. Similar to the Rains relative en... 091
Marco Cerezo @mvscerezo.bsky.social · 22/01/2026I don’t fully know what this duality means yet in a deeper conceptual sense, and I’d genuinely love to hear how others interpret it. Is it “just” representation-theoretic bookkeeping… or is it telling us something more structural about free vs. typical states ? Thoughts welcomed!! 030
Marco Cerezo @mvscerezo.bsky.social · 22/01/2026And then there’s the part that sounds like I’m pretending to be wise at a blackboard: we uncover an “s-duality” relating the phase-space spectra of free states and typical Haar-random (very resourceful) states via a shift in s. 120
Marco Cerezo @mvscerezo.bsky.social · 22/01/2026s = −1 → low-pass / smoothing (keeps the easy/free symmetry content, washes out the spiky, resourceful stuff) s = 0 → no filter (the spectrum is carried over as-is) s = +1 → high-pass / edge detector (suppresses the “free” modes and highlights the highly resourceful ones) 110
Marco Cerezo @mvscerezo.bsky.social · 22/01/2026The punchline: quantum phase space isn’t just a pretty picture of your quantum state… it’s basically a signal-processing pipeline with an actual “EQ knob”. We show that the Stratonovich–Weyl phase-space (tuned by the Cahill–Glauber parameter s) acts like a group-Fourier filter: 110
Marco Cerezo @mvscerezo.bsky.social · 22/01/2026I am very excited to share my latest paper from the quantum computing summer arxiv.org/abs/2601.14225 In collaboration w/ Luke Coffman, @nahuelldiaz.bsky.social , @martinlaroo.bsky.social , and the one and only Maria Shuld from Xanaduarxiv.orgGroup Fourier filtering of quantum resources in quantum phase spaceRecently, it has been shown that group Fourier analysis of quantum states, i.e., decomposing them into the irreducible representations (irreps) of a symmetry group, enables new ways to characterize th... 1123
Marco Cerezo @mvscerezo.bsky.social · 22/12/2025🎆As the year ends, I want to remind everyone that our summer school applications are open! Please apply/encourage your students to apply here: academicjobsonline.org/ajo/jobs/31108 We also have a handshake link too🤝 : app.joinhandshake.com/emp/jobs/105... Shares appreciated!academicjobsonline.orgLos Alamos National Laboratory, Condensed Matter and Complex Systems Job #AJO31108, 2026 Los Alamos Quantum Computing Summer School, Condensed Matter and Complex Systems, Los Alamos National Laboratory, Los Alamos, New Mexico, US 084
Marco Cerezo @mvscerezo.bsky.social · 21/11/2025I am extremely proud of this review article to Quantum Machine Learning 🌟🌟 arxiv.org/abs/2511.15969 Our goal was to give a bird's eye view of QML, from quantum PAC, to Linear Algebraic approaches and variational methods. Huge thanks to my co author Su Yeon Chang@arxiv.orgA Primer on Quantum Machine LearningQuantum machine learning (QML) is a computational paradigm that seeks to apply quantum-mechanical resources to solve learning problems. As such, the goal of this framework is to leverage quantum proce... 0222
Marco Cerezo @mvscerezo.bsky.social · 21/11/2025This was a really cool project to work on! Matt is an amazing researcher, go follow him! Also, he's on the job market for a postdoc 😋😉 051
Reposted by Marco CerezoZoe Holmes @qzoeholmes.bsky.social · 18/11/2025Applications for the LANL quantum computing summer school are now open! I recently got asked about the most important turning points of my career and without hesistation said the LANL summer school. The first 1/3 of this video is me rambling about why it was so special: youtu.be/XjkHmtr_IT0?...youtu.bePrix Zonta 2025 EPFL : Zoë HolmesYouTube video by B&Co Communication 1131
Marco Cerezo @mvscerezo.bsky.social · 18/11/2025🚨Applications for LANL's 2026 Quantum Computing Summer School are open! Please apply here 👇 academicjobsonline.org/ajo/jobs/31108 Reposts appreciated! Deadline to Apply is January 11th 2026. Still, you should go and apply now!academicjobsonline.orgLos Alamos National Laboratory, Condensed Matter and Complex Systems Job #AJO31108, 2026 Los Alamos Quantum Computing Summer School, Condensed Matter and Complex Systems, Los Alamos National Laboratory, Los Alamos, New Mexico, US 1135