Overview
I tested whether quantum kernel SVMs (via PennyLane) actually beat classical PCA+SVC baselines for probing transformer representations. They did not (98.0% vs. 98.7%, p=1.0) — another clean null result, and a useful one to have on record.
Project
Q-TIP
I tested whether quantum kernel SVMs (via PennyLane) actually beat classical PCA+SVC baselines for probing transformer representations. They did not (98.0% vs. 98.7%, p=1.0) — another clean null result, and a useful one to have on record.
Quantum MLInterpretability