diff --git a/avalanche/benchmarks/scenarios/dataset_scenario.py b/avalanche/benchmarks/scenarios/dataset_scenario.py
index 32fb446d3..b5af85058 100644
--- a/avalanche/benchmarks/scenarios/dataset_scenario.py
+++ b/avalanche/benchmarks/scenarios/dataset_scenario.py
@@ -210,7 +210,7 @@ def split_validation_class_balanced(
     # shuffle exp_indices
     exp_indices_t = torch.as_tensor(exp_indices)[torch.randperm(len(exp_indices))]
     # shuffle the targets as well
-    exp_targets = targets_as_tensor[exp_indices]
+    exp_targets = targets_as_tensor[exp_indices_t]
 
     train_exp_indices: list[int] = []
     valid_exp_indices: list[int] = []
diff --git a/tests/benchmarks/scenarios/test_dataset_scenario.py b/tests/benchmarks/scenarios/test_dataset_scenario.py
index 1d2303429..d5c11eabd 100644
--- a/tests/benchmarks/scenarios/test_dataset_scenario.py
+++ b/tests/benchmarks/scenarios/test_dataset_scenario.py
@@ -120,6 +120,7 @@ def test_split_dataset_class_balanced(self):
         for cid in exp.classes_in_this_experience:
             train_cnt = (torch.as_tensor(train_d.targets) == cid).sum()
             valid_cnt = (torch.as_tensor(valid_d.targets) == cid).sum()
+            # print(train_cnt, valid_cnt)
             assert abs(train_cnt - valid_cnt) <= 1
 
         ratio = 0.123
diff --git a/tests/benchmarks/utils/test_avalanche_dataset.py b/tests/benchmarks/utils/test_avalanche_dataset.py
index 92bf5f849..8138e1063 100644
--- a/tests/benchmarks/utils/test_avalanche_dataset.py
+++ b/tests/benchmarks/utils/test_avalanche_dataset.py
@@ -148,7 +148,7 @@ def test_subset_subset_merge(self):
             self.assertTrue(torch.equal(x_curr, x_true))
 
             t_curr = torch.tensor(
-                [curr_dataset.task_labels[idx] for idx in range(d_sz)]
+                [curr_dataset.targets_task_labels[idx] for idx in range(d_sz)]
             )
             t_true = torch.stack([dadata[idx] for idx in true_indices], dim=0)
             self.assertTrue(torch.equal(t_curr, t_true))
diff --git a/tests/training/test_online_strategies.py b/tests/training/test_online_strategies.py
index 9768e67f4..520a837d2 100644
--- a/tests/training/test_online_strategies.py
+++ b/tests/training/test_online_strategies.py
@@ -45,7 +45,7 @@ def test_naive(self):
         benchmark_streams = benchmark.streams.values()
 
         # With task boundaries
-        model, optimizer, criterion, my_nc_benchmark = self.init_sit()
+        model, optimizer, criterion, _ = self.init_sit()
         strategy = OnlineNaive(
             model,
             optimizer,
@@ -113,7 +113,7 @@ def run_strategy_no_boundaries(self, benchmark, cl_strategy):
         cl_strategy.evaluator.loggers = [TextLogger(sys.stdout)]
         results = []
 
-        cl_strategy.train(benchmark.train_stream, num_workers=0)
+        cl_strategy.train(benchmark.train_online_stream, num_workers=0)
         print("Training completed")
 
         assert cl_strategy.clock.train_exp_counter > 0