diff --git a/python/tests/test_params.py b/python/tests/test_params.py
index a51913896..49229104b 100644
--- a/python/tests/test_params.py
+++ b/python/tests/test_params.py
@@ -152,7 +152,7 @@ class TestIVFIndexParam:
     def test_default(self):
         param = IVFIndexParam()
         assert param.metric_type == MetricType.IP
-        assert param.n_list == 0
+        assert param.n_list == 10
         assert param.quantize_type == QuantizeType.UNDEFINED
         assert param.type == IndexType.IVF
 
diff --git a/python/zvec/model/param/__init__.pyi b/python/zvec/model/param/__init__.pyi
index b408866f7..f4ff22aa4 100644
--- a/python/zvec/model/param/__init__.pyi
+++ b/python/zvec/model/param/__init__.pyi
@@ -443,8 +443,7 @@ class IVFIndexParam(VectorIndexParam):
         metric_type (MetricType): Distance metric used for similarity computation.
             Default is ``MetricType.IP`` (inner product).
         n_list (int): Number of clusters (inverted lists) to partition the dataset into.
-            If set to 0, the system will auto-select a reasonable value based on data size.
-            Default is 0 (auto).
+            Default is 10.
         n_iters (int): Number of iterations for k-means clustering during index training.
             Higher values yield more stable centroids. Default is 10.
         use_soar (bool): Whether to enable SOAR (Scalable Optimized Adaptive Routing)
@@ -469,7 +468,7 @@ class IVFIndexParam(VectorIndexParam):
     def __init__(
         self,
         metric_type: _zvec.typing.MetricType = ...,
-        n_list: typing.SupportsInt = 0,
+        n_list: typing.SupportsInt = 10,
         n_iters: typing.SupportsInt = 10,
         use_soar: bool = False,
         quantize_type: _zvec.typing.QuantizeType = ...,
@@ -479,8 +478,8 @@ class IVFIndexParam(VectorIndexParam):
 
         Args:
             metric_type (MetricType, optional): Distance metric. Defaults to MetricType.IP.
-            n_list (int, optional): Number of inverted lists (clusters). Set to 0 for auto.
-                Defaults to 0.
+            n_list (int, optional): Number of inverted lists (clusters).
+                Defaults to 10.
             n_iters (int, optional): Number of k-means iterations during training.
                 Defaults to 10.
             use_soar (bool, optional): Enable SOAR optimization. Defaults to False.
@@ -504,7 +503,7 @@ class IVFIndexParam(VectorIndexParam):
     @property
     def n_list(self) -> int:
         """
-        int: Number of inverted lists (0 = auto).
+        int: Number of inverted lists.
         """
 
     @property
diff --git a/src/binding/python/model/param/python_param.cc b/src/binding/python/model/param/python_param.cc
index 76b7074a5..1045e2324 100644
--- a/src/binding/python/model/param/python_param.cc
+++ b/src/binding/python/model/param/python_param.cc
@@ -794,8 +794,7 @@ and accuracy.
     metric_type (MetricType): Distance metric used for similarity computation.
         Default is ``MetricType.IP`` (inner product).
     n_list (int): Number of clusters (inverted lists) to partition the dataset into.
-        If set to 0, the system will auto-select a reasonable value based on data size.
-        Default is 0 (auto).
+        Default is 10.
     n_iters (int): Number of iterations for k-means clustering during index training.
         Higher values yield more stable centroids. Default is 10.
     use_soar (bool): Whether to enable SOAR (Scalable Optimized Adaptive Routing)
@@ -817,7 +816,7 @@ and accuracy.
 )pbdoc");
   ivf_params
       .def(py::init<MetricType, int, int, bool, QuantizeType>(),
-           py::arg("metric_type") = MetricType::IP, py::arg("n_list") = 0,
+           py::arg("metric_type") = MetricType::IP, py::arg("n_list") = 10,
            py::arg("n_iters") = 10, py::arg("use_soar") = false,
            py::arg("quantize_type") = QuantizeType::UNDEFINED,
            R"pbdoc(
@@ -825,8 +824,8 @@ Constructs an IVFIndexParam instance.
 
 Args:
     metric_type (MetricType, optional): Distance metric. Defaults to MetricType.IP.
-    n_list (int, optional): Number of inverted lists (clusters). Set to 0 for auto.
-        Defaults to 0.
+    n_list (int, optional): Number of inverted lists (clusters).
+        Defaults to 10.
     n_iters (int, optional): Number of k-means iterations during training.
         Defaults to 10.
     use_soar (bool, optional): Enable SOAR optimization. Defaults to False.
@@ -834,7 +833,7 @@ Constructs an IVFIndexParam instance.
         Defaults to QuantizeType.UNDEFINED.
 )pbdoc")
       .def_property_readonly("n_list", &IVFIndexParams::n_list,
-                             "int: Number of inverted lists (0 = auto).")
+                             "int: Number of inverted lists.")
       .def_property_readonly(
           "n_iters", &IVFIndexParams::n_iters,
           "int: Number of k-means iterations during training.")
diff --git a/tests/db/crash_recovery/CMakeLists.txt b/tests/db/crash_recovery/CMakeLists.txt
index 7a872435f..76df6fc28 100644
--- a/tests/db/crash_recovery/CMakeLists.txt
+++ b/tests/db/crash_recovery/CMakeLists.txt
@@ -3,7 +3,8 @@ include(${PROJECT_ROOT_DIR}/cmake/option.cmake)
 
 # Crash recovery tests rely on fork()+execvp()+kill() to spawn helper processes
 # and simulate crashes. These POSIX process APIs are prohibited on iOS.
-if(IOS)
+# On Windows the tests also fail consistently, so skip them there as well.
+if(IOS OR WIN32)
   return()
 endif()