diff --git a/pyproject.toml b/pyproject.toml
index ba8acb64..5c3ad040 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -18,7 +18,7 @@ dependencies = [
     "array-api-compat>=1.13",
     "array-api-extra>=0.10",
     "optree>=0.19",
-    "orng>=0.2.0",
+    "orng>=0.2.1",
 ]
 dynamic = [
     "version"
diff --git a/src/sequince/steps/mutation/metropolis.py b/src/sequince/steps/mutation/metropolis.py
index eb50b451..28e1bc9a 100644
--- a/src/sequince/steps/mutation/metropolis.py
+++ b/src/sequince/steps/mutation/metropolis.py
@@ -6,7 +6,7 @@
 import array_api_extra as xpx
 
 from ..._typing import Array
-from ...array_api import asarray, namespace, to_numpy
+from ...array_api import asarray, get_device, namespace, to_numpy
 from ...target import EvaluationCounts
 from ...transforms import ParticleTransform, select_transform_auxiliary
 from ..base import validate_n_steps
@@ -168,7 +168,9 @@ def apply_with_mutation_state(
             )
             empty_chain = (
                 xp.empty(
-                    (0, *state.particles.shape), dtype=state.particles.dtype
+                    (0, *state.particles.shape),
+                    dtype=state.particles.dtype,
+                    device=get_device(state.particles),
                 )
                 if self._store_chains
                 else None
@@ -531,7 +533,9 @@ def apply_with_mutation_state(
             )
             empty_chain = (
                 xp.empty(
-                    (0, *state.particles.shape), dtype=state.particles.dtype
+                    (0, *state.particles.shape),
+                    dtype=state.particles.dtype,
+                    device=get_device(state.particles),
                 )
                 if self._store_chains
                 else None
@@ -588,7 +592,9 @@ def apply_with_mutation_state(
         acceptance_by_half: list[tuple[float, float]] = []
         chain: list[Array] = []
         for _ in range(n_steps):
-            accepted_population = xp.zeros((n_particles,), dtype=bool)
+            accepted_population = xp.zeros_like(
+                current_log_prob, dtype=xp.bool
+            )
             group_a, group_b = self._split_indices(
                 n_particles,
                 rng,
diff --git a/src/sequince/transforms.py b/src/sequince/transforms.py
index 48a04d4b..af4109a6 100644
--- a/src/sequince/transforms.py
+++ b/src/sequince/transforms.py
@@ -331,8 +331,10 @@ def _log_det_entries(value, arr, xp):
         A vector of log-determinant entries, one per particle.
     """
     if arr.ndim == 1 or arr.shape[0] == 1:
-        return xp.asarray(value)
-    return xp.full((arr.shape[0],), value, dtype=arr.dtype)
+        return asarray(value, xp=xp, like=arr)
+    # Broadcast on the particle device without extracting a host scalar.
+    value = asarray(value, xp=xp, dtype=arr.dtype, like=arr)
+    return xp.broadcast_to(value, (arr.shape[0],))
 
 
 class ParticleTransform: