What exactly does focal loss down-weight?

Question: For a correctly classified example with target-class probability pₜ close to one, what is the primary effect of the focal-loss factor (1−pₜ)^γ with γ>0? A) It reduces that easy example's contribution relative to difficult examples B) It increases its gradient so that confidence approaches one more quickly C) It converts cross-entropy into a margin-ranking objective D) It assigns the example a larger class-balancing coefficient regardless of α Correct: A Explanation: As pₜ approaches one, the modulating factor approaches zero. The loss contribution of an easy, correctly classified example is therefore suppressed, allowing harder examples to have greater influence. Topic: advanced ML / imbalanced learning / focal loss