diff --git a/libcxx/include/__cxx03/__math/traits.h b/libcxx/include/__cxx03/__math/traits.h
index 0d27680d579a4..2b6ca406ae9a0 100644
--- a/libcxx/include/__cxx03/__math/traits.h
+++ b/libcxx/include/__cxx03/__math/traits.h
@@ -12,7 +12,6 @@
 #include <__cxx03/__config>
 #include <__cxx03/__type_traits/enable_if.h>
 #include <__cxx03/__type_traits/is_arithmetic.h>
-#include <__cxx03/__type_traits/is_floating_point.h>
 #include <__cxx03/__type_traits/is_integral.h>
 #include <__cxx03/__type_traits/is_signed.h>
 #include <__cxx03/__type_traits/promote.h>
@@ -28,8 +27,21 @@ namespace __math {
 
 // signbit
 
-template <class _A1, __enable_if_t<is_floating_point<_A1>::value, int> = 0>
-_LIBCPP_NODISCARD inline _LIBCPP_HIDE_FROM_ABI bool signbit(_A1 __x) _NOEXCEPT {
+// The universal C runtime (UCRT) in the WinSDK provides floating point overloads
+// for std::signbit(). By defining our overloads as templates, we can work around
+// this issue as templates are less preferred than non-template functions.
+template <class = void>
+_LIBCPP_NODISCARD inline _LIBCPP_HIDE_FROM_ABI bool signbit(float __x) _NOEXCEPT {
+  return __builtin_signbit(__x);
+}
+
+template <class = void>
+_LIBCPP_NODISCARD inline _LIBCPP_HIDE_FROM_ABI bool signbit(double __x) _NOEXCEPT {
+  return __builtin_signbit(__x);
+}
+
+template <class = void>
+_LIBCPP_NODISCARD inline _LIBCPP_HIDE_FROM_ABI bool signbit(long double __x) _NOEXCEPT {
   return __builtin_signbit(__x);
 }
 
diff --git a/libcxx/test/std/numerics/c.math/signbit.pass.cpp b/libcxx/test/std/numerics/c.math/signbit.pass.cpp
index 44ce32581cca1..7571ced2e4431 100644
--- a/libcxx/test/std/numerics/c.math/signbit.pass.cpp
+++ b/libcxx/test/std/numerics/c.math/signbit.pass.cpp
@@ -17,8 +17,6 @@
 // GCC warns about signbit comparing `bool_v < 0`, which we're testing
 // ADDITIONAL_COMPILE_FLAGS(gcc): -Wno-bool-compare
 
-// XFAIL: FROZEN-CXX03-HEADERS-FIXME
-
 #include <cassert>
 #include <cmath>
 #include <limits>