Every claim of quantum advantage raises the same question: could a classical computer have done this after all? The best way to answer it is to attack the problem with the strongest classical methods available and see where they fail.
Peaked circuits have become one of the clearest test cases for this kind of work. They are built to be hard for classical machines but easy to verify, and a growing number of researchers now try to break peaked circuits classically to test how strong those hardness claims really are. This article walks through how that works, which methods are used, and what the results tell us.
A quick refresher on peaked circuits
A peaked circuit is a quantum circuit designed so one output bitstring appears far more often than any other. In a normal random circuit, probability spreads thinly across a huge number of outcomes and no single result stands out. In a peaked circuit, one hidden result, the peak, has a clearly higher probability.
The idea comes from Scott Aaronson and Yuxuan Zhang. The designer knows the secret bitstring, publishes the circuit, and keeps the answer private. A quantum computer runs the circuit, and the most frequent result is compared with the secret. If they match, the task is solved. Checking takes one comparison.
That easy check is what makes peaked circuits useful. It also sets a very clear goal for anyone trying to beat them classically: find the peak without a quantum computer.
Why classical attacks matter
A hardness claim is only as good as the best attack against it. If a quantum team says a circuit is too hard to simulate, the claim means little until classical experts have tried and failed.
Past quantum advantage results show why this matters. Several random circuit sampling experiments were later challenged when classical teams found faster simulation methods. Each time, the claimed gap between quantum and classical got smaller.
Peaked circuits make this back-and-forth cleaner. The target is one bitstring, and success is a yes or no answer. That makes it easy to compare methods, share results, and see exactly where classical simulation stops working.
The main tool: Matrix Product States
The most common classical attack on peaked circuits uses tensor networks, especially Matrix Product States (MPS).
An MPS stores a quantum state as a chain of smaller tensors, one per qubit. How much information the chain can hold is set by a number called the bond dimension, often written as χ. With a small bond dimension, the simulation is fast and uses little memory. With a larger one, it can capture more entanglement but gets much slower and heavier.
This is the key trade-off. A circuit with low entanglement can be simulated with a small χ. A circuit that builds up a lot of entanglement needs a χ that grows very quickly with circuit depth, and at some point the simulation no longer fits in memory or finishes in reasonable time.
For a peaked circuit, the attacker does not need the full output distribution. They only need the peak to stand out in the simulated result. That is an easier target than full simulation, so even a rough MPS approximation can sometimes reveal the answer.
How a classical attack works in practice
A typical attack follows a simple loop:
- Load the circuit and simulate it with MPS at a low bond dimension.
- Check whether one bitstring clearly stands out in the output.
- If not, raise the bond dimension and run again.
- Repeat until the peak appears or the hardware limit is reached.
The interesting number is the bond dimension where the simulation “cracks” and the peak becomes visible. If that value is small, the circuit is easy to break. If it keeps rising beyond what the hardware can handle, the circuit holds up.
Researchers also track how this breaking point changes as circuits get deeper or wider. For regular random circuits, the required bond dimension grows fast and then levels off. Well-built peaked circuits have been shown to push past that level, which suggests their structure makes classical simulation harder, not easier.
Other tricks attackers use
MPS is not the only option. Classical teams also try:
Smart contraction order. Tensor network methods can contract parts of a circuit in different orders. A good order can cut the cost a lot.
Looking for leftover structure. If the peak was created by tuning gate parameters with an optimizer, the circuit may carry patterns that point to the answer. Spotting those patterns can skip most of the simulation work.
Simplifying the circuit. Some gates cancel or merge. Removing them first makes the rest of the simulation cheaper.
This is why the way a peaked circuit is built matters so much. Early constructions were small, around 24 qubits, and relied on parameter tuning, which does not scale and can leave clues behind. Newer constructions add layers of obfuscation, such as gate swaps and rewrites, so the circuit looks like a generic random one while still hiding the peak.
What the results show so far
Classical attacks have had mixed success, which is a healthy sign. Some heuristic peaked circuit designs have been broken with efficient classical methods, showing their hardness was weaker than hoped. Others, including larger circuits run on real hardware with around 56 qubits and thousands of gates, have held up against strong MPS attacks, with classical estimates reaching years of supercomputer time.
Public challenges add more pressure. Teams publish peaked circuits and invite anyone to recover the hidden bitstring classically. Each attempt, successful or not, tells the field more about where the true limits are.
Why you should try it yourself
Breaking a small peaked circuit on a simulator is one of the best ways to understand quantum simulation. It shows in a very hands-on way:
- how entanglement drives simulation cost
- why bond dimension is the number that matters most
- how circuit design choices change classical hardness
- where the line between “easy” and “hard” actually sits
You do not need a quantum computer to start. A cloud simulator, a few small circuits, and an MPS method are enough to watch a peak appear as the bond dimension rises, and to see the point where larger circuits stop giving in.
Final thoughts
Peaked circuits give quantum computing something it has long been missing: a hard problem with an answer anyone can check. But that only works if classical methods are pushed as hard as possible against them.
Every classical attack, whether it succeeds or fails, makes the picture clearer. Broken designs get replaced by stronger ones, and circuits that survive earn real trust. For researchers and developers alike, trying to break peaked circuits classically is one of the most direct ways to see where the quantum advantage line sits today.
