Cut Noise.
Work Quantum.

This paper demonstrates a high-speed, continuous-variable semi-device-independent quantum random number generator (semi-DI QRNG) proven secure against general adversarial attacks using Kato's inequality and experimental realization.

For cryptographic and machine learning applications, professionals require fast and highly secure true random number generators. This work offers a robust and practical solution by relaxing the strenuous device characterization requirements common in quantum protocols, while offering a strong security guarantee against the sophisticated, non-i.i.d. general attacks that threaten current semi-DI QRNGs. The use of heterodyne detection for simple phase compensation also increases system stability and practicality.