Can intergroup contact improve relations between refugees and host communities? If so, are there added returns to combining contact and empathy education? Does either approach unlock spillover effects among household members? To answer these questions, we conduct a field experiment that brings together Syrian refugees and Lebanese nationals in three localities in Lebanon, where refugees make up a quarter of the population. Working with a Lebanese NGO, we randomly assign Lebanese and Syrian youth participants to an ethnically heterogeneous or homogeneous classroom for a 12-week psycho-social support program. We further randomize whether participants received additional empathy education or a placebo curriculum focused on health and nutrition. We find that contact significantly reduces prejudicial attitudes toward the out-group --- but also depresses participation in future contact, such as attending events celebrating the outgroup's culture 1-2 months after treatment, and worsens mental health among refugee parents and youth. By contrast, we find that empathy education is just as effective --- if not more so --- at reducing prejudicial attitudes, but without these backlash effects. We do not find clear interaction effects of contact and empathy training --- positive effects of the combined treatment are similar to those of empathy alone. The results point to empathy education as a cheaper, more scalable, and potentially even more effective strategy for building social cohesion than intergroup contact.