How does λ affect Generalized Advantage Estimation?

Question: Holding the value-function approximation and discount factor fixed, what is the usual effect of increasing λ in Generalized Advantage Estimation? A) It increases bootstrap dependence and therefore typically lowers variance while increasing bias B) It removes the value-function baseline from the policy-gradient estimator C) It changes the policy-gradient objective from on-policy to off-policy D) It gives greater weight to longer-horizon temporal-difference residuals, usually decreasing bias while increasing variance Correct: D Explanation: Smaller λ emphasises short-horizon bootstrapped estimates, which tend to have lower variance but depend more strongly on value-function accuracy. Larger λ approaches a longer-return or Monte-Carlo-style estimator, usually reducing bootstrap bias while increasing variance. Topic: advanced ML / reinforcement learning / GAE