diff --git a/docs/BENCHMARKS.md b/docs/BENCHMARKS.md
index f5b627e..6a0e21c 100644
--- a/docs/BENCHMARKS.md
+++ b/docs/BENCHMARKS.md
@@ -40,21 +40,21 @@ dotnet run -c Release --project tests/GenDI.Benchmarks/GenDI.Benchmarks.csproj -
## Latest CI benchmark snapshot
-_Updated by [CI run #225](https://github.com/schivei/GenDI/actions/runs/29263183905) on 2026-07-13 15:43 UTC_
+_Updated by [CI run #231](https://github.com/schivei/GenDI/actions/runs/33533719964) on 2026-09-01 16:49 UTC_
| Method | Mean | Allocated |
|---|---:|---:|
-| Manual registration (no GenDI) | 3.211 μs | 7.42 KB |
-| GenDI: constructor injection (generated) | 11.781 μs | 9.98 KB |
-| GenDI: property injection (generated) | 11.780 μs | 9.98 KB |
-| GenDI: with decorator, property injection (generated) | 19.599 μs | 14.26 KB |
-| Reflection registration (no GenDI, assembly scan) | 76.778 μs | 23.9 KB |
+| Manual registration (no GenDI) | 3.206 μs | 7.42 KB |
+| GenDI: constructor injection (generated) | 13.325 μs | 9.98 KB |
+| GenDI: property injection (generated) | 13.287 μs | 9.98 KB |
+| GenDI: with decorator, property injection (generated) | 19.594 μs | 14.26 KB |
+| Reflection registration (no GenDI, assembly scan) | 77.926 μs | 23.9 KB |
### CI analysis
-- GenDI constructor injection is **+266.9%** versus manual registration.
-- GenDI property injection is **+266.9%** versus manual registration.
-- Reflection scanning remains the outlier at **~23.9x slower** and **~3.2x higher allocation** than manual registration.
+- GenDI constructor injection is **+315.6%** versus manual registration.
+- GenDI property injection is **+314.4%** versus manual registration.
+- Reflection scanning remains the outlier at **~24.3x slower** and **~3.2x higher allocation** than manual registration.
- Compatibility note: this benchmark compares manual and generated registrations against a reflection scanner baseline; as documented below, reflection scanning is not suitable for trimming/NativeAOT scenarios, while manual and GenDI-generated registrations remain the supported path.
diff --git a/src/GenDI.Analyzers/GenDI.Analyzers.csproj b/src/GenDI.Analyzers/GenDI.Analyzers.csproj
index d3880e9..75e516e 100644
--- a/src/GenDI.Analyzers/GenDI.Analyzers.csproj
+++ b/src/GenDI.Analyzers/GenDI.Analyzers.csproj
@@ -14,10 +14,10 @@
-
+
diff --git a/src/GenDI.SourceGenerator/GenDI.SourceGenerator.csproj b/src/GenDI.SourceGenerator/GenDI.SourceGenerator.csproj
index 01f4a08..e7ac727 100644
--- a/src/GenDI.SourceGenerator/GenDI.SourceGenerator.csproj
+++ b/src/GenDI.SourceGenerator/GenDI.SourceGenerator.csproj
@@ -16,7 +16,7 @@
-
+
diff --git a/tests/GenDI.Analyzers.Tests/GenDI.Analyzers.Tests.csproj b/tests/GenDI.Analyzers.Tests/GenDI.Analyzers.Tests.csproj
index 209eed5..044a8f3 100644
--- a/tests/GenDI.Analyzers.Tests/GenDI.Analyzers.Tests.csproj
+++ b/tests/GenDI.Analyzers.Tests/GenDI.Analyzers.Tests.csproj
@@ -13,10 +13,10 @@
runtime; build; native; contentfiles; analyzers; buildtransitive
all
-
+
diff --git a/tests/GenDI.SourceGenerator.CoverageDisabled.Tests/GenDI.SourceGenerator.CoverageDisabled.Tests.csproj b/tests/GenDI.SourceGenerator.CoverageDisabled.Tests/GenDI.SourceGenerator.CoverageDisabled.Tests.csproj
index 28e10be..01412b4 100644
--- a/tests/GenDI.SourceGenerator.CoverageDisabled.Tests/GenDI.SourceGenerator.CoverageDisabled.Tests.csproj
+++ b/tests/GenDI.SourceGenerator.CoverageDisabled.Tests/GenDI.SourceGenerator.CoverageDisabled.Tests.csproj
@@ -14,7 +14,7 @@
runtime; build; native; contentfiles; analyzers; buildtransitive
all
-
+
runtime; build; native; contentfiles; analyzers; buildtransitive
all
-
+
runtime; build; native; contentfiles; analyzers; buildtransitive
all
-
+
-_Updated by [CI run #225](https://github.com/schivei/GenDI/actions/runs/29263183905) on 2026-07-13 15:43 UTC_
+_Updated by [CI run #231](https://github.com/schivei/GenDI/actions/runs/33533719964) on 2026-09-01 16:49 UTC_
| Method | Mean | Allocated |
|---|---:|---:|
-| Manual registration (no GenDI) | 3.211 μs | 7.42 KB |
-| GenDI: constructor injection (generated) | 11.781 μs | 9.98 KB |
-| GenDI: property injection (generated) | 11.780 μs | 9.98 KB |
-| GenDI: with decorator, property injection (generated) | 19.599 μs | 14.26 KB |
-| Reflection registration (no GenDI, assembly scan) | 76.778 μs | 23.9 KB |
+| Manual registration (no GenDI) | 3.206 μs | 7.42 KB |
+| GenDI: constructor injection (generated) | 13.325 μs | 9.98 KB |
+| GenDI: property injection (generated) | 13.287 μs | 9.98 KB |
+| GenDI: with decorator, property injection (generated) | 19.594 μs | 14.26 KB |
+| Reflection registration (no GenDI, assembly scan) | 77.926 μs | 23.9 KB |
### CI analysis
-- GenDI constructor injection is **+266.9%** versus manual registration.
-- GenDI property injection is **+266.9%** versus manual registration.
-- Reflection scanning remains the outlier at **~23.9x slower** and **~3.2x higher allocation** than manual registration.
+- GenDI constructor injection is **+315.6%** versus manual registration.
+- GenDI property injection is **+314.4%** versus manual registration.
+- Reflection scanning remains the outlier at **~24.3x slower** and **~3.2x higher allocation** than manual registration.
- Compatibility note: this benchmark compares manual and generated registrations against a reflection scanner baseline; as documented below, reflection scanning is not suitable for trimming/NativeAOT scenarios, while manual and GenDI-generated registrations remain the supported path.