diff --git a/DESCRIPTION b/DESCRIPTION
index c904532..52ba3ac 100644
--- a/DESCRIPTION
+++ b/DESCRIPTION
@@ -23,7 +23,7 @@ Imports:
     checkmate,
     data.table,
     lgr,
-    mlr3 (>= 0.17.0),
+    mlr3 (>= 0.19.0),
     mlr3misc,
     uuid
 Suggests:
diff --git a/NEWS.md b/NEWS.md
index 093c464..8d11f80 100644
--- a/NEWS.md
+++ b/NEWS.md
@@ -3,6 +3,7 @@
 * feat: `reduceResultsBatchmark` gains argument `fun` which is passed on to `batchtools::reduceResultsList`, useful for deleting model data to avoid running out of memory, https://github.com/mlr-org/mlr3batchmark/issues/18 Thanks to Toby Dylan Hocking @tdhock for the PR.
 * docs: A warning is now given when the loaded mlr3 version differs from the
 mlr3 version stored in the trained learners
+* Support marshaling
 
 # mlr3batchmark 0.1.1
 
diff --git a/R/reduceResultsBatchmark.R b/R/reduceResultsBatchmark.R
index 7ba85e8..441c1f2 100644
--- a/R/reduceResultsBatchmark.R
+++ b/R/reduceResultsBatchmark.R
@@ -14,7 +14,8 @@
 #'
 #' @return [mlr3::BenchmarkResult].
 #' @export
-reduceResultsBatchmark = function(ids = NULL, store_backends = TRUE, reg = batchtools::getDefaultRegistry(), fun=NULL) { # nolint
+reduceResultsBatchmark = function(ids = NULL, store_backends = TRUE, reg = batchtools::getDefaultRegistry(), fun = NULL, unmarshal = TRUE) { # nolint
+  assert_flag(unmarshal)
   if (is.null(ids)) {
     ids = batchtools::findDone(ids, reg = reg)
   } else {
@@ -87,5 +88,9 @@ reduceResultsBatchmark = function(ids = NULL, store_backends = TRUE, reg = batch
     bmr$combine(mlr3::BenchmarkResult$new(rdata))
   }
 
+  if (unmarshal) {
+    bmr$unmarshal()
+  }
+
   return(bmr)
 }
diff --git a/R/worker.R b/R/worker.R
index 600d414..613587c 100644
--- a/R/worker.R
+++ b/R/worker.R
@@ -9,6 +9,7 @@ run_learner = function(job, data, learner_hash, store_models, ...) {
     learner = learner,
     resampling = resampling,
     store_models = store_models,
-    lgr_threshold = lgr::get_logger("mlr3")$threshold
+    lgr_threshold = lgr::get_logger("mlr3")$threshold,
+    is_sequential = FALSE
   )
 }
diff --git a/man/reduceResultsBatchmark.Rd b/man/reduceResultsBatchmark.Rd
index d1964e1..7eb04e2 100644
--- a/man/reduceResultsBatchmark.Rd
+++ b/man/reduceResultsBatchmark.Rd
@@ -8,7 +8,8 @@ reduceResultsBatchmark(
   ids = NULL,
   store_backends = TRUE,
   reg = batchtools::getDefaultRegistry(),
-  fun = NULL
+  fun = NULL,
+  unmarshal = TRUE
 )
 }
 \arguments{
@@ -34,6 +35,12 @@ Registry. If not explicitly passed, uses the default registry (see \code{\link[b
 \item{fun}{[\code{function}]\cr
 Function to apply to each result. The result is passed unnamed as first argument. If \code{NULL}, the identity is used.
 If the function has the formal argument \dQuote{job}, the \code{\link[batchtools]{Job}}/\code{\link[batchtools]{Experiment}} is also passed to the function.}
+
+\item{unmarshal}{\code{\link[mlr3]{Learner}}\cr
+Whether to unmarshal learners that were marshaled during the execution.
+Setting this to \code{FALSE} does not guarantee that the learners are marshaled.
+For example, with sequential execution and no encapsulation, marshaling is not necessary.
+If you want to ensure that all learners are in marshaled form, you need to call \verb{$marshal()} on the result object.}
 }
 \value{
 \link[mlr3:BenchmarkResult]{mlr3::BenchmarkResult}.
diff --git a/tests/testthat/test_batchmark.R b/tests/testthat/test_batchmark.R
index 86b0153..b00e11e 100644
--- a/tests/testthat/test_batchmark.R
+++ b/tests/testthat/test_batchmark.R
@@ -86,3 +86,14 @@ test_that("failing jobs", {
   expect_data_table(as.data.table(results), nrow = 8L)
   expect_error(reduceResultsBatchmark(reg = reg, ids = ids), "successfully computed")
 })
+
+test_that("marshaling", {
+  reg = batchtools::makeExperimentRegistry(NA)
+  batchmark(benchmark_grid(tsk("iris"), lrn("classif.debug"), rsmp("holdout")), store_models = TRUE)
+  submitJobs()
+  bmr_unmarshaled = reduceResultsBatchmark(unmarshal = TRUE)
+  bmr_marshaled = reduceResultsBatchmark(unmarshal = FALSE)
+
+  expect_true(bmr_marshaled$resample_result(1)$learners[[1]]$marshaled)
+  expect_false(bmr_unmarshaled$resample_result(1)$learners[[1]]$marshaled)
+})