Not All Benchmarks Are Created Equal

on YouTube ↗01h 03m

" summary

A survey argues that benchmarking needs variety and clarity: a 1,500-source bibliography catalogs methods, testbeds run at 5–50 qubits, and many benchmarks probe different properties rather than a single “goodness” number. Tools compared include state tomography, randomized benchmarking and direct RB scales to ~10 qubits, Google’s random-circuit sampling/Cross-Entropy on ~53 qubits, IBM’s quantum volume and a proposed volumetric benchmarks framework; experiments show circuit structure and compilation change outcomes and efficiently verifiable circuits are needed next.

AI-written summary · about

" chapters (15)

1. 0:00Start
2. 11:25Characterization Questions
3. 21:49Future Digital Quantum Computers
4. 22:09Analog Simulators
5. 25:34General Guidelines
6. 31:58Characterizing Elementary Operations
7. 41:38Direct Randomized Benchmarking
8. 44:50The Significance of Using Clifford Circuits
9. 45:19Why Do You Use Clifford Circuits
10. 46:59Quantum Supremacy
11. 49:55Quantum Volume
12. 52:46Random Circuit Sampling
13. 54:15Volumetric Benchmarks
14. 56:56Conclusion
15. 58:47Protocols for Holistic Benchmarking