ETH-Hardness of Approximating 2-CSPs and Directed Steiner Network

We study the 2-ary constraint satisfaction problems (2-CSPs), which can be stated as follows: given a constraint graph $G=(V,E)$, an alphabet set $\Sigma$ and, for each $\{u, v\}\in E$, a constraint $C_{uv} \subseteq \Sigma\times\Sigma$, the goal is to find an assignment $\sigma: V \to \Sigma$ that satisfies as many constraints as possible, where a constraint $C_{uv}$ is satisfied if $(\sigma(u),\sigma(v))\in C_{uv}$. While the approximability of 2-CSPs is quite well understood when $|\Sigma|$ is constant, many problems are still open when $|\Sigma|$ becomes super constant. One such problem is whether it is hard to approximate 2-CSPs to within a polynomial factor of $|\Sigma| |V|$. Bellare et al. (1993) suggested that the answer to this question might be positive. Alas, despite efforts to resolve this conjecture, it remains open to this day. In this work, we separate $|V|$ and $|\Sigma|$ and ask a related but weaker question: is it hard to approximate 2-CSPs to within a polynomial factor of $|V|$ (while $|\Sigma|$ may be super-polynomial in $|V|$)? Assuming the exponential time hypothesis (ETH), we answer this question positively by showing that no polynomial time algorithm can approximate 2-CSPs to within a factor of $|V|^{1 - o(1)}$. Note that our ratio is almost linear, which is almost optimal as a trivial algorithm gives a $|V|$-approximation for 2-CSPs. Thanks to a known reduction, our result implies an ETH-hardness of approximating Directed Steiner Network with ratio $k^{1/4 - o(1)}$ where $k$ is the number of demand pairs. The ratio is roughly the square root of the best known ratio achieved by polynomial time algorithms (Chekuri et al., 2011; Feldman et al., 2012). Additionally, under Gap-ETH, our reduction for 2-CSPs not only rules out polynomial time algorithms, but also FPT algorithms parameterized by $|V|$. Similar statement applies for DSN parameterized by $k$.