db={token:[{"token_id":1},{"token_id":2},{"token_id":2},{"token_id":3},{"token_id":4},{"token_id":5},{"token_id":5}]}

db.token.aggregate([{$group:{_id:"$token_id",/** Get the first document for each unique field*/doc:{$first:"$$ROOT"}}},{$replaceRoot:{/** Replace root to get back the original document structure*/newRoot:"$doc"}},{$sort:{token_id:1}},{$skip:3},{$limit:2}])