diff --git a/resolve/likelihood.py b/resolve/likelihood.py
index 54c9ed462895fd652a9118a32abd05cbb5cd7967..642c71597fb9155453de1ae3944b58aa69f7e72e 100644
--- a/resolve/likelihood.py
+++ b/resolve/likelihood.py
@@ -55,7 +55,7 @@ def _build_gauss_lh_nres(op, mean, invcov):
     my_assert_isinstance(mean, invcov, (ift.Field, ift.MultiField))
     my_asserteq(op.target, mean.domain, invcov.domain)
 
-    lh = ift.GaussianEnergy(mean=mean, inverse_covariance=ift.makeOp(invcov)) @ op
+    lh = ift.GaussianEnergy(mean=mean, inverse_covariance=ift.makeOp(invcov, sampling_dtype=mean.dtype)) @ op
     return _Likelihood(lh, mean, lambda x: ift.makeOp(invcov), op)
 
 
diff --git a/resolve/minimization.py b/resolve/minimization.py
index 32ef0a7207e3886ff278bf1ae867b495ebffe6d1..dbf004c5b0dc23bfedbe40316820f3e1ef0b9bc0 100644
--- a/resolve/minimization.py
+++ b/resolve/minimization.py
@@ -28,7 +28,7 @@ class Minimization:
         else:
             my_assert(n_samples > 0)
             dct = {
-                "mean": position,
+                "position": position,
                 "hamiltonian": operator,
                 "n_samples": n_samples,
                 "minimizer_sampling": None,
@@ -38,7 +38,7 @@ class Minimization:
                 "comm": comm,
                 "nanisinf": True,
             }
-            self._e = ift.SampledKL(**dct)
+            self._e = ift.SampledKLEnergy(**dct)
             self._n, self._m = dct["n_samples"], dct["mirror_samples"]
 
     def minimize(self, minimizer):