diff --git a/docs/src/content/docs/benchmarks/dupradar.md b/docs/src/content/docs/benchmarks/dupradar.mdx
similarity index 69%
rename from docs/src/content/docs/benchmarks/dupradar.md
rename to docs/src/content/docs/benchmarks/dupradar.mdx
index 570c445..dc5b900 100644
--- a/docs/src/content/docs/benchmarks/dupradar.md
+++ b/docs/src/content/docs/benchmarks/dupradar.mdx
@@ -3,6 +3,8 @@ title: dupRadar Benchmark
 description: Performance comparison between RustQC and R dupRadar, with zero output mismatches across all values.
 ---
 
+export const base = import.meta.env.BASE_URL;
+
 RustQC produces output **identical** to the R dupRadar package. Every value in the duplication matrix, every count, and the fitted model parameters match exactly. The benchmarks below quantify the performance difference on real RNA-seq data.
 
 ## Small benchmark
@@ -69,39 +71,39 @@ The plots below compare the R dupRadar output (left) with RustQC output (right)
 
 #### Density scatter plot
 
-<div style="display: grid; grid-template-columns: 1fr 1fr; gap: 1rem; margin: 1rem 0;">
-  <figure style="margin: 0; text-align: center;">
-    <img src="/benchmarks/large/dupRadar/duprateExpDens.png" alt="dupRadar (R) density scatter plot" style="width: 100%;" />
+<div style={{display: "grid", gridTemplateColumns: "1fr 1fr", gap: "1rem", margin: "1rem 0"}}>
+  <figure style={{margin: 0, textAlign: "center"}}>
+    <img src={`${base}benchmarks/large/dupRadar/duprateExpDens.png`} alt="dupRadar (R) density scatter plot" style={{width: "100%"}} />
     <figcaption>dupRadar (R)</figcaption>
   </figure>
-  <figure style="margin: 0; text-align: center;">
-    <img src="/benchmarks/large/RustQC/GM12878_REP1.markdup.sorted_duprateExpDens.png" alt="RustQC density scatter plot" style="width: 100%;" />
+  <figure style={{margin: 0, textAlign: "center"}}>
+    <img src={`${base}benchmarks/large/RustQC/GM12878_REP1.markdup.sorted_duprateExpDens.png`} alt="RustQC density scatter plot" style={{width: "100%"}} />
     <figcaption>RustQC</figcaption>
   </figure>
 </div>
 
 #### Duplication rate boxplot
 
-<div style="display: grid; grid-template-columns: 1fr 1fr; gap: 1rem; margin: 1rem 0;">
-  <figure style="margin: 0; text-align: center;">
-    <img src="/benchmarks/large/dupRadar/duprateExpBoxplot.png" alt="dupRadar (R) duplication rate boxplot" style="width: 100%;" />
+<div style={{display: "grid", gridTemplateColumns: "1fr 1fr", gap: "1rem", margin: "1rem 0"}}>
+  <figure style={{margin: 0, textAlign: "center"}}>
+    <img src={`${base}benchmarks/large/dupRadar/duprateExpBoxplot.png`} alt="dupRadar (R) duplication rate boxplot" style={{width: "100%"}} />
     <figcaption>dupRadar (R)</figcaption>
   </figure>
-  <figure style="margin: 0; text-align: center;">
-    <img src="/benchmarks/large/RustQC/GM12878_REP1.markdup.sorted_duprateExpBoxplot.png" alt="RustQC duplication rate boxplot" style="width: 100%;" />
+  <figure style={{margin: 0, textAlign: "center"}}>
+    <img src={`${base}benchmarks/large/RustQC/GM12878_REP1.markdup.sorted_duprateExpBoxplot.png`} alt="RustQC duplication rate boxplot" style={{width: "100%"}} />
     <figcaption>RustQC</figcaption>
   </figure>
 </div>
 
 #### Expression histogram
 
-<div style="display: grid; grid-template-columns: 1fr 1fr; gap: 1rem; margin: 1rem 0;">
-  <figure style="margin: 0; text-align: center;">
-    <img src="/benchmarks/large/dupRadar/expressionHist.png" alt="dupRadar (R) expression histogram" style="width: 100%;" />
+<div style={{display: "grid", gridTemplateColumns: "1fr 1fr", gap: "1rem", margin: "1rem 0"}}>
+  <figure style={{margin: 0, textAlign: "center"}}>
+    <img src={`${base}benchmarks/large/dupRadar/expressionHist.png`} alt="dupRadar (R) expression histogram" style={{width: "100%"}} />
     <figcaption>dupRadar (R)</figcaption>
   </figure>
-  <figure style="margin: 0; text-align: center;">
-    <img src="/benchmarks/large/RustQC/GM12878_REP1.markdup.sorted_expressionHist.png" alt="RustQC expression histogram" style="width: 100%;" />
+  <figure style={{margin: 0, textAlign: "center"}}>
+    <img src={`${base}benchmarks/large/RustQC/GM12878_REP1.markdup.sorted_expressionHist.png`} alt="RustQC expression histogram" style={{width: "100%"}} />
     <figcaption>RustQC</figcaption>
   </figure>
 </div>
diff --git a/docs/src/content/docs/benchmarks/featurecounts.md b/docs/src/content/docs/benchmarks/featurecounts.mdx
similarity index 51%
rename from docs/src/content/docs/benchmarks/featurecounts.md
rename to docs/src/content/docs/benchmarks/featurecounts.mdx
index fcc1181..b36c84a 100644
--- a/docs/src/content/docs/benchmarks/featurecounts.md
+++ b/docs/src/content/docs/benchmarks/featurecounts.mdx
@@ -3,6 +3,8 @@ title: featureCounts Benchmark
 description: Performance and accuracy comparison between RustQC's built-in read counting and Rsubread featureCounts.
 ---
 
+export const base = import.meta.env.BASE_URL;
+
 RustQC includes built-in gene-level read counting that produces output
 compatible with the Subread featureCounts format. This page compares the
 counting performance and output accuracy against standalone Rsubread
@@ -48,6 +50,49 @@ exactly.
 The output format is directly compatible with downstream tools such as DESeq2
 and MultiQC.
 
+### Side-by-side plots
+
+Since dupRadar visualizations are derived from the same gene-level counts, identical plots confirm that the underlying feature counting is correct. The plots below compare R dupRadar output (left) with RustQC output (right) for the large benchmark.
+
+#### Density scatter plot
+
+<div style={{display: "grid", gridTemplateColumns: "1fr 1fr", gap: "1rem", margin: "1rem 0"}}>
+  <figure style={{margin: 0, textAlign: "center"}}>
+    <img src={`${base}benchmarks/large/dupRadar/duprateExpDens.png`} alt="dupRadar (R) density scatter plot" style={{width: "100%"}} />
+    <figcaption>dupRadar (R)</figcaption>
+  </figure>
+  <figure style={{margin: 0, textAlign: "center"}}>
+    <img src={`${base}benchmarks/large/RustQC/GM12878_REP1.markdup.sorted_duprateExpDens.png`} alt="RustQC density scatter plot" style={{width: "100%"}} />
+    <figcaption>RustQC</figcaption>
+  </figure>
+</div>
+
+#### Duplication rate boxplot
+
+<div style={{display: "grid", gridTemplateColumns: "1fr 1fr", gap: "1rem", margin: "1rem 0"}}>
+  <figure style={{margin: 0, textAlign: "center"}}>
+    <img src={`${base}benchmarks/large/dupRadar/duprateExpBoxplot.png`} alt="dupRadar (R) duplication rate boxplot" style={{width: "100%"}} />
+    <figcaption>dupRadar (R)</figcaption>
+  </figure>
+  <figure style={{margin: 0, textAlign: "center"}}>
+    <img src={`${base}benchmarks/large/RustQC/GM12878_REP1.markdup.sorted_duprateExpBoxplot.png`} alt="RustQC duplication rate boxplot" style={{width: "100%"}} />
+    <figcaption>RustQC</figcaption>
+  </figure>
+</div>
+
+#### Expression histogram
+
+<div style={{display: "grid", gridTemplateColumns: "1fr 1fr", gap: "1rem", margin: "1rem 0"}}>
+  <figure style={{margin: 0, textAlign: "center"}}>
+    <img src={`${base}benchmarks/large/dupRadar/expressionHist.png`} alt="dupRadar (R) expression histogram" style={{width: "100%"}} />
+    <figcaption>dupRadar (R)</figcaption>
+  </figure>
+  <figure style={{margin: 0, textAlign: "center"}}>
+    <img src={`${base}benchmarks/large/RustQC/GM12878_REP1.markdup.sorted_expressionHist.png`} alt="RustQC expression histogram" style={{width: "100%"}} />
+    <figcaption>RustQC</figcaption>
+  </figure>
+</div>
+
 ## Additional outputs
 
 Beyond the standard featureCounts counts file and summary, RustQC also
diff --git a/docs/src/content/docs/outputs/dupradar.mdx b/docs/src/content/docs/outputs/dupradar.mdx
index a5057e5..dde1499 100644
--- a/docs/src/content/docs/outputs/dupradar.mdx
+++ b/docs/src/content/docs/outputs/dupradar.mdx
@@ -5,6 +5,8 @@ description: Comprehensive guide to RustQC's dupRadar output files, including th
 
 import { Card, CardGrid, FileTree } from '@astrojs/starlight/components';
 
+export const base = import.meta.env.BASE_URL;
+
 ## What is dupRadar?
 
 In DNA sequencing, PCR duplicates are straightforward to identify and remove: reads with identical alignment coordinates are likely amplification artifacts. In **RNA-seq**, the situation is more nuanced.
@@ -118,15 +120,15 @@ See the [Interpreting plots](#interpreting-plots) section below for how to read
 <CardGrid>
   <Card title="Density scatter plot">
     `<sample>_duprateExpDens.{png,svg}`
-    ![Density scatter plot](/RustQC/examples/dupradar/duprateExpDens.png)
+    <img src={`${base}examples/dupradar/duprateExpDens.png`} alt="Density scatter plot" />
   </Card>
   <Card title="Duplication rate boxplot">
     `<sample>_duprateExpBoxplot.{png,svg}`
-    ![Duplication rate boxplot](/RustQC/examples/dupradar/duprateExpBoxplot.png)
+    <img src={`${base}examples/dupradar/duprateExpBoxplot.png`} alt="Duplication rate boxplot" />
   </Card>
   <Card title="Expression histogram">
     `<sample>_expressionHist.{png,svg}`
-    ![Expression histogram](/RustQC/examples/dupradar/expressionHist.png)
+    <img src={`${base}examples/dupradar/expressionHist.png`} alt="Expression histogram" />
   </Card>
 </CardGrid>
 
diff --git a/docs/src/content/docs/outputs/rseqc.mdx b/docs/src/content/docs/outputs/rseqc.mdx
index f4bd3a2..19a6776 100644
--- a/docs/src/content/docs/outputs/rseqc.mdx
+++ b/docs/src/content/docs/outputs/rseqc.mdx
@@ -5,6 +5,8 @@ description: Output files produced by RustQC's RSeQC-compatible quality control
 
 import { Card, CardGrid, FileTree } from '@astrojs/starlight/components';
 
+export const base = import.meta.env.BASE_URL;
+
 RustQC reimplements seven tools from the [RSeQC](https://rseqc.sourceforge.net/)
 package. Each tool produces output files that match the format and content of the
 original Python implementation, plus **native PNG and SVG plots** generated
@@ -87,7 +89,7 @@ high duplication.
 
 <CardGrid>
   <Card title="Duplication rate">
-    <img src="/RustQC/examples/rseqc/read_duplication.png" alt="Read duplication rate plot" />
+    <img src={`${base}examples/rseqc/read_duplication.png`} alt="Read duplication rate plot" />
   </Card>
 </CardGrid>
 
@@ -136,10 +138,10 @@ known (blue), partial novel (red), and complete novel (green) splicing.
 
 <CardGrid>
   <Card title="Splice events">
-    <img src="/RustQC/examples/rseqc/splice_events.png" alt="Splice events pie chart" />
+    <img src={`${base}examples/rseqc/splice_events.png`} alt="Splice events pie chart" />
   </Card>
   <Card title="Splice junctions">
-    <img src="/RustQC/examples/rseqc/splice_junction.png" alt="Splice junctions pie chart" />
+    <img src={`${base}examples/rseqc/splice_junction.png`} alt="Splice junctions pie chart" />
   </Card>
 </CardGrid>
 
@@ -169,7 +171,7 @@ detected (in thousands). A plateau in the curves indicates saturation.
 
 <CardGrid>
   <Card title="Junction saturation">
-    <img src="/RustQC/examples/rseqc/junction_saturation.png" alt="Junction saturation plot" />
+    <img src={`${base}examples/rseqc/junction_saturation.png`} alt="Junction saturation plot" />
   </Card>
 </CardGrid>
 
@@ -210,7 +212,7 @@ approximately normal for a well-prepared library.
 
 <CardGrid>
   <Card title="Inner distance">
-    <img src="/RustQC/examples/rseqc/inner_distance.png" alt="Inner distance histogram" />
+    <img src={`${base}examples/rseqc/inner_distance.png`} alt="Inner distance histogram" />
   </Card>
 </CardGrid>