LeetCode 1696 Jump Game VI - Optimization

Look at the transition: $\max_{j=i-k}^{i-1}(dp[j])$. This is exactly the sliding window maximum problem! The window contains the last $k$ DP values.

You already know how to solve that in $O(1)$ per query using a monotonic decreasing deque. Apply it directly to the DP values. As you compute $dp[i]$, maintain a deque of indices sorted by decreasing $dp$ values. The front gives you the maximum $dp$ value in the valid range.