In Evo-GDPO, raw task performance, latency, and monetary cost are combined into a single scalar reward function.

In the primary objective formulation $R_i^{\text{rew}} = r_i + \lambda_{\text{evo}} [r_i - b_r]_+$, what does the parameter $\lambda_{\text{evo}}$ control?

Explain the architectural consequence on latency ($R_i^{\text{lat}}$) and cost ($R_i^{\text{cost}}$) channels when a candidate harness underperforms relative to the incumbent reward baseline ($r_i < b_r$).

Match each baseline statistic from the archive incumbent to the metric it represents in Evo-GDPO.

In the Evo-GDPO evaluation formulation, the unnormalized reward signal $R_i^{\text{rew}}$ serves as the ___ objective.

Order the mathematical operations used to compute the latency channel $R_i^{\text{lat}}$ for a candidate harness.

Decoupled Advantage Normalization in Evo-GDPO