diff --git a/R-package/R/xgb.cv.R b/R-package/R/xgb.cv.R
index 71269419b..258c7fb16 100644
--- a/R-package/R/xgb.cv.R
+++ b/R-package/R/xgb.cv.R
@@ -95,7 +95,18 @@ xgb.cv <- function(params=list(), data, nrounds, nfold, label = NULL, missing =
   }
 
   folds <- xgb.cv.mknfold(dtrain, nfold, params)
-  predictValues <- rep(0,xgb.numrow(dtrain))
+  obj_type = params[['objective']]
+  mat_pred = FALSE
+  if (!is.null(obj_type) && obj_type=='multi:softprob')
+  {
+    num_class = params[['num_class']]
+    if (is.null(num_class))
+      stop('must set num_class to use softmax')
+    predictValues <- matrix(0,xgb.numrow(dtrain),num_class)
+    mat_pred = TRUE
+  }
+  else
+    predictValues <- rep(0,xgb.numrow(dtrain))
   history <- c()
   for (i in 1:nrounds) {
     msg <- list()
@@ -106,7 +117,10 @@ xgb.cv <- function(params=list(), data, nrounds, nfold, label = NULL, missing =
         msg[[k]] <- xgb.iter.eval(fd$booster, fd$watchlist, i - 1, feval) %>% str_split("\t") %>% .[[1]]
       } else {
         res <- xgb.iter.eval(fd$booster, fd$watchlist, i - 1, feval, prediction)
-        predictValues[fd$index] <- res[[2]]
+        if (mat_pred)
+          predictValues[fd$index,] <- res[[2]]
+        else
+          predictValues[fd$index] <- res[[2]]
         msg[[k]] <- res[[1]] %>% str_split("\t") %>% .[[1]]
       }
     }