improved tab styling

app/src/components/Tabs.astro

@@ -1,9 +1,16 @@
 ---
-const { class: className, ...props } = Astro.props;
+interface Props {
+  title?: string;
+  class?: string;
+  [key: string]: any;
+}
+const { title, class: className, ...props } = Astro.props;
 const wrapperClass = ["tabs", className].filter(Boolean).join(" ");
 ---
 <div class={wrapperClass} {...props}>
-  <div class="tabs__nav" role="tablist"></div>
+  <div class="tabs__nav" role="tablist">
+    {title && <span class="tabs__title">{title}</span>}
+  </div>
   <div class="tabs__panels">
     <slot />
   </div>
@@ -82,6 +89,20 @@ const wrapperClass = ["tabs", className].filter(Boolean).join(" ");
     display: none;
   }
 
+  .tabs__title {
+    padding: var(--spacing-2) var(--spacing-3);
+    border-right: 1px solid var(--border-color);
+    font-size: 0.8em;
+    font-weight: 700;
+    color: var(--text-muted);
+    text-transform: uppercase;
+    letter-spacing: 0.05em;
+    white-space: nowrap;
+    display: flex;
+    align-items: center;
+    user-select: none;
+  }
+
   .tabs__nav :global(.tabs__btn) {
     flex: 1 1 0;
     padding: var(--spacing-2) var(--spacing-3);

app/src/content/chapters/2-setup.mdx

@@ -31,7 +31,7 @@ For inference we use vLLM [@vllm] with tensor parallelism, chunked prefill, and
 
 Before diving into experiments, here's a quick overview of the datasets we compare against. We use "source data" and "seed data" interchangeably throughout.
 
-<Tabs>
+<Tabs title="Curated">
 <Tab title="DCLM">
   A standardized benchmark providing a 240T token corpus from Common Crawl with model-based filtering as a key curation strategy. DCLM (DataComp-LM) enables training a 7B parameter model to 64% accuracy on MMLU with 2.6T tokens [@datacomp].
 </Tab>
@@ -41,6 +41,9 @@ Before diving into experiments, here's a quick overview of the datasets we compa
 <Tab title="Ultra-FineWeb">
   A 1T English token and 120B Chinese token dataset created by applying efficient verification-based filtering to FineWeb. Uses a lightweight fastText classifier and optimized seed data selection to improve data quality [@ultrafineweb].
 </Tab>
+</Tabs>
+
+<Tabs title="Synthetic">
 <Tab title="Nemotron-HQ-Synth">
   Part of Nemotron-CC, a 6.3T token dataset using classifier ensembling and synthetic data rephrasing. The High-Quality-Synthetic subset contains synthetically rephrased data using Qwen3-30B-A3B [@qwen3] [@nemotroncc].
 </Tab>