Fidelity
Fidelity is a mainstream metric in the XAI field for measuring explanation quality, divided into local and global fidelity. It measures the degree to which an explanation reproduces model outputs, but cannot guarantee that the explanation faithfully reflects the underlying decision process.
Timeline (last 90 days)
Infidelity is 1 minus fidelity, the complement of fidelity, and in scenarios where infidelity is extremely small it is approximately equal to the square of the trace distance between the prepared state and the target state.
Fidelity is a measure quantifying the closeness between quantum states; for a pure state |ψ⟩ and a target state |φ⟩, fidelity simplifies to |⟨ψ|φ⟩|²; for a mixed state ρ and a target pure state |φ⟩, it is ⟨φ|ρ|φ⟩.
The current mainstream approach in academia for evaluating explanation quality is to measure the fidelity between explanations and black-box model predictions.
The paper points out that fidelity can only quantify whether an explanation can reproduce model outputs, but cannot guarantee that the explanation correctly reflects the underlying decision-making process.
All Facts (4)
Infidelity is 1 minus fidelity, the complement of fidelity, and in scenarios where infidelity is extremely small it is approximately equal to the square of the trace distance between the prepared state and the target state.
50%UnverifiedFidelity is a measure quantifying the closeness between quantum states; for a pure state |ψ⟩ and a target state |φ⟩, fidelity simplifies to |⟨ψ|φ⟩|²; for a mixed state ρ and a target pure state |φ⟩, it is ⟨φ|ρ|φ⟩.
50%UnverifiedThe current mainstream approach in academia for evaluating explanation quality is to measure the fidelity between explanations and black-box model predictions.
50%UnverifiedThe paper points out that fidelity can only quantify whether an explanation can reproduce model outputs, but cannot guarantee that the explanation correctly reflects the underlying decision-making process.
50%