The primary means I've seen it happen by is by being part of a team that does pair programming rather than code reviews. In general I've seen it do wonders for team alignment, integration speed, and ramping up new colleagues, junior or otherwise. It does take a concerted effort to learn and do well though. There are best practices, and putting two random programmers behind the same computer does not necessarily mean they'll find and apply those.

I'm curious to see if there's been studies on its efficacy, and whether or not those studies agree with my entirely anecdotal experience.