Updated
Updated · New Scientist · Aug 17
Analysis Finds 73% of 4,966 Quantum Computing Papers Cannot Be Replicated
Updated
Updated · New Scientist · Aug 17

Analysis Finds 73% of 4,966 Quantum Computing Papers Cannot Be Replicated

1 articles · Updated · New Scientist · Aug 17

Summary

  • A review of 4,966 quantum-computing papers found only 26.8% included enough information to even attempt replication, raising doubts about many reported advances in the field.
  • In a manual check of 127 papers, just 24.4% provided runnable code, and 64.5% of those code packages failed to execute successfully on independent systems.
  • Researchers said the problem reflects both weak reproducibility practices and the instability of still-maturing quantum hardware, whose changing physical characteristics make results hard to reproduce across machines.
  • Some experts said the findings are not alarming for a young field, arguing innovation still outranks reproducible infrastructure, while others expect community standards and AI coding tools to improve replication over time.

Insights

As quantum hardware crashes and APIs shift, will strict reproducibility rules save the industry or stifle its greatest innovations?
If most quantum breakthroughs fail to run twice, are we investing billions into a mirage of unstable hardware?