Can one tool be most accurate for everyone?

No. Accuracy leadership changes by use case.

What is the best evaluation metric?

False-positive cost plus false-negative cost in your workflow.

How to evaluate 'most accurate' claims with your own benchmark set and practical error tolerance.

The phrase 'most accurate' is context-dependent. Accuracy changes by text type and review threshold.

Test tools on your own data, then compare consistency and error costs.