diff --git a/pyproject.toml b/pyproject.toml
index b468c7e6..ace4846a 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -5,7 +5,7 @@ build-backend = "scikit_build_core.build"
 
 [project]
 name = "pyvale"
-version = "2026.10.0"
+version = "2026.10.1"
 description = "Your virtual engineering lab: An all-in-one package for simulation-driven experimental design."
 authors = [
   { name = "scepticalrabbit et al.", email = "thescepticalrabbit@gmail.com" },
diff --git a/src/pyvale/__init__.py b/src/pyvale/__init__.py
index a56656d5..cd7bbe12 100644
--- a/src/pyvale/__init__.py
+++ b/src/pyvale/__init__.py
@@ -14,7 +14,7 @@
 virtually testing digital shadows/twins.
 """
 
-__version__ = "2026.9.0"
+__version__ = "2026.10.1"
 
 from . import data, dataio, dic, mooseherder, render, sensorsim, strain
 
diff --git a/src/pyvale/dic/cpp/bindings.cpp b/src/pyvale/dic/cpp/bindings.cpp
index 578cd95a..be6db4fe 100644
--- a/src/pyvale/dic/cpp/bindings.cpp
+++ b/src/pyvale/dic/cpp/bindings.cpp
@@ -15,6 +15,7 @@
 #include <cstring>
 #include <stdexcept>
 #include <type_traits>
+#include <algorithm>
 
 // commoncpp Header Files
 #include "../../commoncpp/util.hpp"
@@ -174,6 +175,22 @@ PYBIND11_MODULE(diccpp, m) {
         engine_images(images_from_sequence(image_arrays, conf), img_roi_arr, calib, conf, multiwindowconf, saveconf);
     }, "Run DIC analysis on in-memory image arrays with config");
 
+    m.def("_should_update_ref", [](int image_number,
+                                    const std::vector<int>& iterations,
+                                    const std::vector<double>& costs,
+                                    util::EIncrementalCond condition,
+                                    double threshold) {
+        const int num_results = static_cast<int>(std::max(iterations.size(), costs.size()));
+        ResultArrays results(num_results, 0, false);
+        results.niter = iterations;
+        results.cost = costs;
+
+        util::Config config{};
+        config.incremental_update_cond = condition;
+        config.incremental_update_val = threshold;
+        return should_update_ref(image_number, results, config);
+    }, "Evaluate an incremental-reference update condition");
+
     // interpolator bindings
     py::class_<InterpVals>(m, "InterpVals")
         .def_readonly("f", &InterpVals::f)
diff --git a/src/pyvale/dic/cpp/dicmain.cpp b/src/pyvale/dic/cpp/dicmain.cpp
index 044e8d3e..2d3d2fd9 100644
--- a/src/pyvale/dic/cpp/dicmain.cpp
+++ b/src/pyvale/dic/cpp/dicmain.cpp
@@ -573,7 +573,7 @@ bool should_update_ref(const int img_num_def_l, const ResultArrays& results, con
             if (results.cost.empty()) return false;
 
             double avg = std::accumulate(results.cost.begin(), results.cost.end(), 0.0) / results.cost.size();
-            return avg > conf.incremental_update_val;
+            return avg < conf.incremental_update_val;
         }
 
         default:
diff --git a/tests/dic/test_incremental.py b/tests/dic/test_incremental.py
new file mode 100644
index 00000000..f66efeff
--- /dev/null
+++ b/tests/dic/test_incremental.py
@@ -0,0 +1,40 @@
+"""Tests for incremental DIC reference-update conditions."""
+
+import pytest
+
+import pyvale.dic.diccpp as diccpp
+
+
+@pytest.mark.parametrize(
+    ("condition", "image_number", "iterations", "costs", "threshold", "expected"),
+    [
+        # The first deformed image never updates the reference.
+        ("IMAGE", 1, [4], [0.5], 1, False),
+        ("IMAGE", 2, [4], [0.5], 1, True),
+        ("IMAGE", 3, [4], [0.5], 2, True),
+        # COST updates only when the mean ZNCC cost is strictly below the threshold.
+        ("COST", 2, [4, 4], [0.4, 0.6], 0.6, True),
+        ("COST", 2, [4, 4], [0.4, 0.6], 0.5, False),
+        ("COST", 2, [4, 4], [0.4, 0.6], 0.4, False),
+        # ITER updates only when the mean iteration count exceeds the threshold.
+        ("ITER", 2, [4, 6], [0.5, 0.5], 4.9, True),
+        ("ITER", 2, [4, 6], [0.5, 0.5], 5, False),
+    ],
+)
+def test_incremental_update_conditions(
+    condition: str,
+    image_number: int,
+    iterations: list[int],
+    costs: list[float],
+    threshold: float,
+    expected: bool,
+) -> None:
+    result = diccpp._should_update_ref(
+        image_number,
+        iterations,
+        costs,
+        getattr(diccpp.IncrementalCond, condition),
+        threshold,
+    )
+
+    assert result is expected