diff --git a/examples/modeling/arad.py b/examples/modeling/arad.py
index 20d124bd..60eb7023 100644
--- a/examples/modeling/arad.py
+++ b/examples/modeling/arad.py
@@ -7,6 +7,13 @@
 import tensorflow as tf
 from riid.models.neural_nets.arad import ARAD, ARADv1TF, ARADv2TF
 
+import numpy as np
+
+from riid.data.synthetic import get_dummy_seeds
+from riid.data.synthetic.seed import SeedMixer
+from riid.data.synthetic.static import StaticSynthesizer
+
+
 if len(sys.argv) == 2:
     import matplotlib
     matplotlib.use("Agg")
@@ -33,8 +40,44 @@ def show_summaries(model):
         pass
 
 
-arad_v1 = ARAD(model=ARADv1TF())
+arad_v1 = ARAD(model=ARADv1TF(latent_dim=5))
 show_summaries(arad_v1.model)
 
-arad_v2 = ARAD(model=ARADv2TF())
+arad_v2 = ARAD(model=ARADv2TF(latent_dim=8))
 show_summaries(arad_v2.model)
+
+# Generate some training data
+fg_seeds_ss, bg_seeds_ss = get_dummy_seeds().split_fg_and_bg()
+mixed_bg_seed_ss = SeedMixer(bg_seeds_ss, mixture_size=3).generate(1)
+
+static_synth = StaticSynthesizer(
+    samples_per_seed=250,
+    snr_function="log10",
+    return_fg=False,
+    return_gross=True,
+)
+_, train_ss = static_synth.generate(fg_seeds_ss, mixed_bg_seed_ss)
+train_ss.normalize()
+
+print("training ARADv1...")
+arad_v1.fit(train_ss, epochs=50, verbose=True)
+print("training ARADv2...")
+arad_v2.fit(train_ss, epochs=50, verbose=True)
+
+# Generate some test data
+static_synth.samples_per_seed = 50
+_, test_ss = static_synth.generate(fg_seeds_ss, mixed_bg_seed_ss)
+test_ss.normalize()
+
+# Predict
+arad_v1_reconstructions = arad_v1.predict(test_ss, verbose=True)
+recon_errors = test_ss.info["recon_error"].values
+ood_decisions = test_ss.info["ood"].values
+print((f"ARADv1: mean reconstruction error = {np.mean(recon_errors):.3f} (KLD)\n"
+       f"                         OOD rate = {np.mean(ood_decisions):.2f}"))
+
+arad_v2_reconstructions = arad_v2.predict(test_ss, verbose=True)
+recon_errors = test_ss.info["recon_error"].values
+ood_decisions = test_ss.info["ood"].values
+print((f"ARADv2: mean reconstruction error = {np.mean(recon_errors):.3f} (JSD)\n"
+       f"                         OOD rate = {np.mean(ood_decisions):.2f}"))
diff --git a/riid/models/neural_nets/arad.py b/riid/models/neural_nets/arad.py
index 10e47824..a9bf725f 100644
--- a/riid/models/neural_nets/arad.py
+++ b/riid/models/neural_nets/arad.py
@@ -89,7 +89,11 @@ def __init__(self, latent_dim: int = 5):
         decoder_input = Input(shape=(latent_dim,), name="decoder_input")
         x = Dense(units=40, name="D2")(decoder_input)
         x = Dropout(rate=0.1, name="D2_dropout")(x)
-        decoder_output = Dense(units=128, name="D3")(x)
+        decoder_output = Dense(
+            units=128,
+            activation=sigmoid,  # unclear from paper, seems to be necessary
+            name="D3"
+        )(x)
         decoder = Model(decoder_input, decoder_output, name="decoder")
 
         # Autoencoder
@@ -129,7 +133,7 @@ class ARADv2TF(Model):
     - Ghawaly Jr, James M., et al. "Characterization of the Autoencoder Radiation Anomaly Detection
       (ARAD) model." Engineering Applications of Artificial Intelligence 111 (2022): 104761.
     """
-    def __init__(self, latent_dim: int = 5):
+    def __init__(self, latent_dim: int = 8):
         """
         Args:
             latent_dim: dimension of internal latent represention.