Demonstration notebook for RCS and XEB analysis by shashwatk1998 · Pull Request #469 · quantumlib/ReCirq

shashwatk1998 · 2026-03-30T22:34:36Z

No description provided.

This configures Gemini Code Assist to be a little bit less noisy and verbose. Based on experiences with other repos, I find this makes Gemini more useful.

And make it bit more apparent how is `RCSResults._fidelities_lin` set. Fixes mypy recirq/random_circuit_sampling | grep random_circuit

Co-authored-by: Pavol Juhas <pavol.juhas@gmail.com>

review-notebook-app · 2026-03-30T22:34:41Z

Check out this pull request on

See visual diffs & provide feedback on Jupyter Notebooks.

Powered by ReviewNB

deltorobarba

Moved comments directly to code positions

deltorobarba · 2026-04-02T17:24:32Z

+      }
+    },
+    {
+      "cell_type": "code",


The metadata in the JSON is configured for Google Colab ("colab": {"provenance": []}). So since this notebook is Colab-friendly, should we add a commented-out '!pip install cirq cirq-google qsimcirq' cell at the top and also a comment on how to install recirq + restarting kernel? It might save external users a few steps if they try to run this directly from the repo.

deltorobarba · 2026-04-02T17:24:53Z

+      "source": [
+        "# Calculate linear XEB fidelities\n",
+        "\n",
+        "To calculate fidelities via noiseless simulation, we can employ the standard `cirq.Simulator()`. For experiments involving larger system sizes, the more performance-optimized `qsimcirq.QSimSimulator()` can be used to improve execution efficiency."


In the "Calculate linear XEB fidelities" markdown cell, you state: "To calculate fidelities via noiseless simulation, we can employ the standard cirq.Simulator()". However, the code cell right after it actually uses qsimcirq.QSimSimulator(). You might just want to slightly tweak the markdown text so it matches your code execution seamlessly (e.g., "Here, we use the performance-optimized qsimcirq.QSimSimulator() to improve execution efficiency.").

Thanks, done!

deltorobarba · 2026-04-02T17:25:17Z

+      "source": [
+        "# Plot the results\n",
+        "\n",
+        "Visualize the XEB results for each patch across the specified circuit depths. The plot reflects the mean fidelity calculated across all random circuit instances for each (patch, depth) configuration."


The plotted fidelity is perfectly flat at around 1.0 because characterize=False and used for default, noiseless cirq.Simulator(). On real quantum hardware experiment (or a noisy simulation), errors would accumulate as the circuit gets deeper. Since this is a tutorial, a beginner who is familiar with real hardware RCS might be confused why there's no exponential decay curve as one would expect on real QC hardware. Would it be worth adding a quick sentence above the plot explicitly stating that we see a flat 1.0 line because this is an ideal, noiseless simulation? Unless you think users immediately understand this behavior for noiseless simulations. Here a proposal: "Note on Ideal vs. Noisy Simulation: You might notice that the plotted linear XEB fidelity is flat at around 1.0 across all circuit depths. This behavior is exactly what we expect from an ideal, noiseless simulation. In a real quantum hardware experiment (or a noisy simulation), incoherent errors and decoherence accumulate at each gate layer, resulting in a characteristic exponential decay curve as the circuit gets deeper. Since our simulated qubits do not suffer from any noise, the fidelity remains at ~1.0 (with microscopic fluctuations resulting purely from finite sampling variance across our 10,000 repetitions)."

@shashwatk1998 maybe it's worth adding results from the noisy simulator too? You could use the QVM here https://quantumai.google/cirq/simulate/qvm_basic_example. If the number of qubits isn't too large, it might run faster with cirq.DensityMatrixSimulator instead of the statevector simulators, which will use quantum trajectories.

Added the note and also added noisy simulation

eliottrosenberg · 2026-04-14T06:53:53Z

+      "cell_type": "code",
+      "source": [
+        "# To run this notebook in Google Colab, uncomment and run the following line:\n",
+        "#!pip install cirq cirq_google qsimcirq\n",


Do you actually use qsimcirq?

Yes, for calculating fidelity in noiseless simulation.

fidelities = results.fidelities_lin(simulator=qsimcirq.QSimSimulator())

eliottrosenberg · 2026-04-14T06:54:23Z

+        "noise_props = cirq_google.engine.load_device_noise_properties(processor_id)\n",
+        "noise_model = cirq_google.NoiseModelFromGoogleNoiseProperties(noise_props)\n",
+        "sim = cirq.DensityMatrixSimulator(noise=noise_model)"


sim is unused

sim is used in the definition of sim_processor

sim_processor = cirq_google.engine.SimulatedLocalProcessor( processor_id=processor_id, sampler=sim, device=device, calibrations={cal.timestamp // 1000: cal} )

eliottrosenberg · 2026-04-14T06:55:10Z

+        ")\n",
+        "sim_engine = cirq_google.engine.SimulatedLocalEngine([sim_processor])\n",
+        "qvm_sampler = sim_engine.get_sampler(processor_id)\n",
+        "print(sim_engine.get_processor(processor_id).get_device())"


This line is a little funny. You could have just done print(device)

Fair. Have been following how QVM is defined: https://quantumai.google/cirq/simulate/quantum_virtual_machine

eliottrosenberg · 2026-04-14T06:55:41Z

+        "cal = cirq_google.engine.load_median_device_calibration(processor_id)\n",
+        "sim_processor = cirq_google.engine.SimulatedLocalProcessor(\n",
+        "    processor_id=processor_id, sampler=sim, device=device, calibrations={cal.timestamp // 1000: cal}\n",
+        ")\n",
+        "sim_engine = cirq_google.engine.SimulatedLocalEngine([sim_processor])\n",
+        "qvm_sampler = sim_engine.get_sampler(processor_id)\n",


Instead of all of these lines, you could just use sim that you got above instead of qvm_sampler

Agreed, I had followed how it is in qvm documentation: https://quantumai.google/cirq/simulate/quantum_virtual_machine

Will update to use sim directly.

eliottrosenberg · 2026-04-14T07:03:41Z

+        "\n",
+        "\n",
+        "ax.set_xlim(min(noisy_experiment.depths) - 2, max(noisy_experiment.depths) + 2)\n",
+        "ax.set_xlabel('Cycle Depth', fontsize=14)\n",


Maybe "Number of cycles, $m$" to match Fig 4b from Nature 574, 505–510 (2019).

Sure, will do!

eliottrosenberg · 2026-04-14T07:05:03Z

+        "# Define the patches\n",
+        "\n",
+        "Select sets of connected qubits to serve as patches. These patches must be disjoint and should not share any common qubits."
+      ],


What is the purpose of the patches? What are they used for in Nature 574, 505–510 (2019)? Give a bit more motivation here.

Thanks, will add description.

eliottrosenberg · 2026-04-14T17:37:05Z

+        "for i in range(len(patches_on_device)):\n",
+        "    x = noisy_experiment.depths\n",
+        "    y = [np.mean(noisy_fidelities[(i, d)]) for d in x]\n",
+        "    ax.plot(x, y, marker='o', markersize=8, linewidth=2, label=f'Patch {i}')\n",


Can we add error bars to this plot and then put it on a log-y scale, restricting to cycle numbers where you can distinguish the fidelity from 0?

Shashwat Kumar and others added 14 commits February 27, 2026 22:58

Add RCS experiment

194915b

typo

c73f29c

fix docstrings

60b82ca

Merge branch 'main' into rcs_v0

7e6e7e3

Add Gemini Code Assist config file (quantumlib#462)

c45382d

This configures Gemini Code Assist to be a little bit less noisy and verbose. Based on experiences with other repos, I find this makes Gemini more useful.

addressed comments

cbc089c

Merge branch 'main' into rcs_v0

4debe3e

Fix incompatible return value found by typecheck

b21fdb8

And make it bit more apparent how is `RCSResults._fidelities_lin` set. Fixes mypy recirq/random_circuit_sampling | grep random_circuit

Update recirq/random_circuit_sampling/rcs_experiment.py

fbb390b

Co-authored-by: Pavol Juhas <pavol.juhas@gmail.com>

Update recirq/random_circuit_sampling/rcs_experiment.py

5eee62e

Co-authored-by: Pavol Juhas <pavol.juhas@gmail.com>

Update recirq/random_circuit_sampling/rcs_experiment.py

d49d465

Co-authored-by: Pavol Juhas <pavol.juhas@gmail.com>

addressed Alexis's comments

d7cb15d

Merge branch 'main' into rcs_v0

35d222a

docs: add demonstration notebook for RCS and XEB analysis

541f349

deltorobarba reviewed Apr 2, 2026

View reviewed changes

Shashwat Kumar added 2 commits April 13, 2026 21:12

update characterize_pairs and add a test for it using ideal fSim gates

3545e2c

addressed the comments on the notebook

5857254

eliottrosenberg reviewed Apr 14, 2026

View reviewed changes

Merge branch 'main' into rcs_v0

f8e7354

Conversation

shashwatk1998 commented Mar 30, 2026

Uh oh!

review-notebook-app bot commented Mar 30, 2026

Uh oh!

deltorobarba left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

deltorobarba left a comment •

edited

Loading