Same Transition, Two Updates

An agent takes action $a$ in state $s$, receives $r = 2$, and lands in $s'$. Current estimates:

$Q(s,a) = 4.0, \qquad Q(s', \text{up}) = 3.0, \quad Q(s', \text{down}) = -1.0, \quad Q(s', \text{right}) = 7.0, \quad Q(s', \text{left}) = 0.0$

The $\varepsilon$-greedy behaviour policy explores at $s'$ and selects down. With $\alpha = 0.5$ and $\gamma = 0.9$, what is $Q(s,a)$ after this transition under Q-learning and under Sarsa?