(I/III) We're excited to announce a new tenure track opening! The position is called 'quantum informatics' and is affiliated with our QUICK group within the CS+AI division at @jku.at 🇦🇹. Application deadline is November 30th, 2025: www.jku.at/en/the-jku/w...
I mean I can actually imagine people trying challenges with NISQ computers down the line.
My student @johnbostanci.bsky.social, Chinmay Nirkhe, Jonas Haferkamp, and Mark Zhandry have put out a tour-de-force paper that shows, relative to a classical oracle, QMA is stronger than QCMA -- i.e., quantum proofs >> classical proofs. Congratulations to the authors! arxiv.org/abs/2511.09551
Grover's quadratic speedup is provably optimal in the black-box setting. We expect general SAT to be essentially unstructured and as hard as the black-box setting (strong exponential time hypothesis). It's believable that Grover is optimal there. But this is not clear for 3-SAT.
Also, thanks to my great collaborators, mxkramer.bsky.social, Alexander Nietner and @jenseisert.bsky.social!
3-SAT is clearly more structured than SAT in general, as evidenced by the improved classical runtime of O(1.307^n) compared to the 2^n for SAT. Using Grover on top of a classical solver, this structure is only addressed classically, inherently capping the achievable speedup at quadratic.
Grover's quadratic speedup is provably optimal in the black-box setting. We expect general SAT to be essentially unstructured and as hard as the black-box setting (strong exponential time hypothesis). It's believable that Grover is optimal there. But this is not clear for 3-SAT.
To be clear: This line of research is not about exponential quantum advantages, but polynomial ones. 3-SAT is an NP-hard problem and is NOT expected to be efficiently solvable on a quantum computer.
To get a worst-case scaling of the form O(c^n), we would need a lower bound on the gap of the SAT instance encoding Hamiltons. Unfortunately, we did not manage this and after talking to a few experts, this seems to be a difficult task. We leave this as an open problem.
Regarding the algorithmic improvements, I particularly like our perfect hash family based scheme for parallelizing local measurements in the algorithm.
Obvious candidate because strong numerical performance is known for this algorithm, but theoretical understanding is limited. We provide an expression for the worst-case runtime depending on the Hamiltonian gap as well as interesting algorithmic improvements.
We look at such an algorithm in scirate.com/arxiv/2511.0..., namely the measurement-driven SAT solver from Benjamin, Zhao and Fitzsimons, which I think is an obvious candidate algorithm for this line of thought.
BZF algorithm: arxiv.org/abs/1711.02687
The fastest quantum algorithms for 3-SAT achieve only quadratic, Grover-type speedups over the best classical algorithms. Optimality of this is unknown. I think we should look at algorithms that do more than put Grover on top of a classical base. (See link for paper below)
Quantum state tomography for states that vary with parameters such as time or control settings attains new capabilities in characterization of evolving quantum states.
go.aps.org/3SAelEa
Artificially intelligent Maxwell's demon for optimal control of open quantum systems
iopscience.iop.org/article/10.1...
A reinforcement learning agent in #machinelearning is interpreted literally as a thermodynamic agent reminiscent of a #Maxwell's demon for the control of open quantum systems.
🎉I am very happy to see our Barren Plateau Review published in Nature Review Physics. This article condenses 6 years of our LANL work, but also so many amazing papers by the community!!
Check it here 👇
www.nature.com/articles/s42...
The "hide the exponential" game Scott Aaronson has complained about for decades is alive and well: arxiv.org/abs/2412.13164
Yes, Shor's algorithm with 3 registers only needs very basic operations.
No, you can't pack a 2^2000 level quantum state into 1 oscillator and then operate accurately on it.
Here is the referee report for that paper. :) Same would apply here.
New work out today 📝
Very insightful collaboration with colleagues from Fraunhofer HHI and the great @jenseisert.bsky.social. We offer a preview of explainable AI #xAI for #Quantum learning models #QML ⚛️🧠. Check it out and let us know your thoughts!
It seems that OpenAI's latest model, o3, can solve 25% of problems on a database called FrontierMath, created by EpochAI, where previous LLMs could only solve 2%. On Twitter I am quoted as saying, "Getting even one question right would be well beyond what we can do now, let alone saturating them."
Very happy with this joint paper with Paula Belzig, Li Gao, and Peixue Wu! The main goal is to understand how much distinguishability can be preserved under the action of a noisy channel. We study this with (relative) expansion coefficients.
scirate.com/arxiv/2411.1...
Troy's morning lecture was truly great, as a friendly introduction to quantum computing for algorithms designers.
Abstracting the quantum aspects as "new rules of the game, and blackbox primitives you can use" is really speaking to a classical (T)CS audience!
The plot thickens:
"While all the proofs in the paper are correct to the best of our knowledge, we have been recently informed about a classical attack on our polynomial system."
Just looked for the code and there is a Julia package, I love that!
The PCP theorem, a jewel of theoretical computer science, establishes that any NP statement can be assessed by a randomized verifier who only checks a vanishing fraction of the proof (indeed, a constant # of characters!)
This has had incredible impact, most notably on how ML reviews are conducted
It's tough to gain visibility as a young researcher, and it's job market season! Are you a theoretical computer science PhD/postdoc on the job market?
I don't have a crazy juge audience but I'll try to help a bit: fill this form, and I'll tweet your pitch and info!
docs.google.com/forms/d/e/1F...
In the spirit of trying to use this thing properly, rather than just lurking like I did on TwiXer, a community I'd like to be more plugged into is people using memory systems for research (Anki, Mochi, etc, but for more than language learning or memorising capital cities). So I've made a list.