diff --git a/botorch/acquisition/logei.py b/botorch/acquisition/logei.py
index e6316deeac..2d95b9f49f 100644
--- a/botorch/acquisition/logei.py
+++ b/botorch/acquisition/logei.py
@@ -18,11 +18,8 @@
 from __future__ import annotations
 
 from collections.abc import Callable
-
 from copy import deepcopy
-
 from functools import partial
-
 from typing import TypeVar
 
 import torch
@@ -215,7 +212,7 @@ def __init__(
             tau_max=check_tau(tau_max, name="tau_max"),
             fat=fat,
         )
-        self.register_buffer("best_f", torch.as_tensor(best_f, dtype=float))
+        self.register_buffer("best_f", torch.as_tensor(best_f))
         self.tau_relu = check_tau(tau_relu, name="tau_relu")
 
     def _sample_forward(self, obj: Tensor) -> Tensor:
diff --git a/botorch/acquisition/monte_carlo.py b/botorch/acquisition/monte_carlo.py
index 9e39fc6db4..c2fe72fafd 100644
--- a/botorch/acquisition/monte_carlo.py
+++ b/botorch/acquisition/monte_carlo.py
@@ -401,7 +401,7 @@ def __init__(
             constraints=constraints,
             eta=eta,
         )
-        self.register_buffer("best_f", torch.as_tensor(best_f, dtype=float))
+        self.register_buffer("best_f", torch.as_tensor(best_f))
 
     def _sample_forward(self, obj: Tensor) -> Tensor:
         r"""Evaluate qExpectedImprovement per sample on the candidate set `X`.
@@ -715,9 +715,9 @@ def __init__(
             constraints=constraints,
             eta=eta,
         )
-        best_f = torch.as_tensor(best_f, dtype=float).unsqueeze(-1)  # adding batch dim
+        best_f = torch.as_tensor(best_f).unsqueeze(-1)  # adding batch dim
         self.register_buffer("best_f", best_f)
-        self.register_buffer("tau", torch.as_tensor(tau, dtype=float))
+        self.register_buffer("tau", torch.as_tensor(tau))
 
     def _sample_forward(self, obj: Tensor) -> Tensor:
         r"""Evaluate qProbabilityOfImprovement per sample on the candidate set `X`.
diff --git a/botorch_community/acquisition/rei.py b/botorch_community/acquisition/rei.py
index 91c7f8dda5..f617e50ce3 100644
--- a/botorch_community/acquisition/rei.py
+++ b/botorch_community/acquisition/rei.py
@@ -187,7 +187,7 @@ def __init__(
         sample_dim = tuple(range(len(self.sample_shape) + 1))
         self._sample_reduction = partial(logmeanexp, dim=sample_dim)
 
-        self.register_buffer("best_f", torch.as_tensor(best_f, dtype=float))
+        self.register_buffer("best_f", torch.as_tensor(best_f))
         self.fat: bool = fat
         self.tau_relu: float = check_tau(tau_relu, "tau_relu")
         dim: int = X_dev.shape[1]
diff --git a/test/acquisition/test_logei.py b/test/acquisition/test_logei.py
index 189df17617..6d5ebb7695 100644
--- a/test/acquisition/test_logei.py
+++ b/test/acquisition/test_logei.py
@@ -46,7 +46,6 @@
 from botorch.sampling.normal import IIDNormalSampler, SobolQMCNormalSampler
 from botorch.utils.low_rank import sample_cached_cholesky
 from botorch.utils.testing import BotorchTestCase, MockModel, MockPosterior
-
 from botorch.utils.transforms import standardize
 from torch import Tensor
 
@@ -131,8 +130,10 @@ def test_q_log_expected_improvement(self):
                 self.assertIn(k, acqf._modules)
                 self.assertIn(k, log_acqf._modules)
 
-            res = acqf(X).item()
-            self.assertEqual(res, 0.0)
+            res = acqf(X)
+            self.assertEqual(res.dtype, dtype)
+            self.assertEqual(res.device.type, self.device.type)
+            self.assertEqual(res.item(), 0.0)
             exp_log_res = log_acqf(X).exp().item()
             # Due to the smooth approximation, the value at zero should be close to, but
             # not exactly zero, and upper-bounded by the tau hyperparameter.