diff --git a/vizier/_src/algorithms/designers/gp/acquisitions.py b/vizier/_src/algorithms/designers/gp/acquisitions.py
index 71add9d6d..56f0cc0bd 100644
--- a/vizier/_src/algorithms/designers/gp/acquisitions.py
+++ b/vizier/_src/algorithms/designers/gp/acquisitions.py
@@ -196,6 +196,26 @@ def __call__(
     return tfp_bo.acquisition.GaussianProcessExpectedImprovement(dist, labels)()
 
 
+@struct.dataclass
+class ExpectedValue(AcquisitionFunction):
+  """Expected Improvement acquisition function."""
+
+  def __call__(
+      self,
+      dist: tfd.Distribution,
+      features: Optional[types.ModelInput] = None,
+      labels: Optional[types.PaddedArray] = None,
+      seed: Optional[jax.random.KeyArray] = None,
+  ) -> jax.Array:
+    del features, seed
+    if labels is not None:
+      labels = labels.replace_fill_value(-np.inf).padded_array
+    return (
+        tfp_bo.acquisition.GaussianProcessExpectedImprovement(dist, labels)()
+        + dist.mean()
+    )
+
+
 @struct.dataclass
 class PI(AcquisitionFunction):
   """Probability of Improvement acquisition function."""
@@ -245,6 +265,16 @@ def default_ucb_pi(cls) -> 'AcquisitionTrustRegion':
         UCB(1.8), PI(), bad_acq_value=-1e12, threshold=0.3, apply_tr_after=0
     )
 
+  @classmethod
+  def default_ucb_expected_value(cls) -> 'AcquisitionTrustRegion':
+    return cls(
+        UCB(1.8),
+        ExpectedValue(),
+        bad_acq_value=-1e12,
+        threshold=None,
+        apply_tr_after=0,
+    )
+
   @classmethod
   def default_ucb_lcb(cls) -> 'AcquisitionTrustRegion':
     return cls(
@@ -289,9 +319,10 @@ def __call__(
     apply_tr = False
     if labels is not None:
       labels_padded = labels.replace_fill_value(np.nan).padded_array
-      threshold = jnp.minimum(
-          jnp.nanmean(labels_padded), jnp.nanmedian(labels_padded)
-      )
+      # threshold = jnp.minimum(
+      #     jnp.nanmean(labels_padded), jnp.nanmedian(labels_padded)
+      # )
+      threshold = jnp.nanmedian(labels_padded)
       apply_tr = labels._original_shape[0] <= self.apply_tr_after
     if self.threshold is not None:
       threshold = self.threshold