1: \begin{abstract}

2: 

3: A celebrated result of H{\aa}stad established that, for any constant $\varepsilon>0$,

4: it is NP-hard to find an assignment satisfying a $(1/|\gr|+\varepsilon)$-fraction

5: of the constraints of a given $\eq$ instance over an Abelian group $\gr$ even if

6: one is promised that an assignment satisfying a $(1-\varepsilon)$-fraction of

7: the constraints exists. Engebretsen, Holmerin, and Russell showed the same

8: result for $\eq$ instances over any finite (not necessarily Abelian) group.

9: In other words, for almost-satisfiable instances of $\eq$ the random

10: assignment achieves an optimal approximation guarantee.

11: %

12: We prove that the random assignment algorithm is still best possible under a stronger

13: promise that the $\eq$ instance is almost satisfiable over an arbitrarily

14: more restrictive group.

15: 

16: \end{abstract}

17: