docs: highlight benchmark dashboard and links

mihailoxyz · mihailoxyz · commit 77809431a887 · 2026-02-26T17:18:29.000+01:00
diff --git a/README.md b/README.md
@@ -4,10 +4,41 @@
 [![Python 3.10+](https://img.shields.io/badge/python-3.10%2B-blue.svg)](https://www.python.org/downloads/)
 [![BixBench-Verified-50](https://img.shields.io/badge/benchmark-BixBench--Verified--50-green.svg)](https://huggingface.co/datasets/phylobio/BixBench-Verified-50)
 [![BioAgents](https://img.shields.io/badge/backend-BioAgents-purple.svg)](https://github.com/bio-xyz/BioAgents)
-[![Get API Key](https://img.shields.io/badge/API%20Key-chat.bio.xyz-teal.svg)](https://chat.bio.xyz/)
+[![Get API Key](https://img.shields.io/badge/API%20Key-chat.bio.xyz-teal.svg)](https://chat.bio.xyz/chat?settings=account&section=api-keys)
 
 BixBench evaluation harness for [BioAgents](https://github.com/bio-xyz/BioAgents) and its closed-source literature and data-analysis agents.
 
+### Overall highlights (from dashboard)
+
+| Mode                    | Accuracy | 95% Wilson CI | Correct / Total |
+| ----------------------- | -------: | ------------: | --------------: |
+| Direct                  |    71.3% | 67.4% - 74.9% |       392 / 550 |
+| MCQ with refusal        |    85.1% | 81.9% - 87.8% |       468 / 550 |
+| **MCQ without refusal** |    90.0% | 87.2% - 92.2% |       495 / 550 |
+
+| Headline                                                  |               Value |
+| --------------------------------------------------------- | ------------------: |
+| MCQ lift (Direct -> **MCQ without refusal**)              |             +18.7pp |
+| Refusal gap (**MCQ without refusal** -> MCQ with refusal) |              -4.9pp |
+| Best repeat (**MCQ without refusal**)                     | 96.0% (`085809-r3`) |
+| Task groups at 100%                                       |     22 / 32 (68.8%) |
+
+### Showcase: Benchmark Dashboard
+
+Interactive page from `docs/`:
+
+- Live (GitHub Pages): [bio-xyz.github.io/bio-benchmark](https://bio-xyz.github.io/bio-benchmark/)
+- Source: [`docs/index.html`](docs/index.html)
+
+[![Benchmark dashboard preview](docs/assets/performance_analysis.png)](https://bio-xyz.github.io/bio-benchmark/)
+
+Local preview:
+
+```bash
+python3 -m http.server 8080 --directory docs
+# open http://localhost:8080
+```
+
 ### Minimal config
 
 Only these fields are required:
@@ -76,7 +107,5 @@ python3 scripts/build_pages_data.py
 This script updates:
 
 - `docs/data/benchmark_summary.json`
-- `docs/data/results/*.csv` (only the 3 scoped CSV files)
+- `docs/data/results/*.csv`
 - `docs/assets/performance_analysis.png`
-
-GitHub Pages deploys automatically via `.github/workflows/pages.yml` on pushes to `main` that touch `docs/` or relevant source result files. In repository settings, set **Pages -> Build and deployment -> Source** to **GitHub Actions**.
diff --git a/docs/index.html b/docs/index.html
@@ -3,24 +3,21 @@
   <head>
     <meta charset="UTF-8" />
     <meta name="viewport" content="width=device-width, initial-scale=1.0" />
-    <title>PhyloBioBixBench-Verified-50 | Bio Benchmark Results</title>
+    <title>BIOS Benchmark Results</title>
     <meta
       name="description"
-      content="Interactive dashboard for PhyloBioBixBench-Verified-50 benchmark results — direct, MCQ, and refusal-aware grading across 50 biology questions."
+      content="Interactive dashboard for BixBench-Verified-50 benchmark results — direct, MCQ, and refusal-aware grading across 50 biology questions."
     />
 
     <!-- Favicon -->
     <link rel="icon" type="image/svg+xml" href="assets/favicon.svg" />
 
     <!-- Open Graph -->
     <meta property="og:type" content="website" />
-    <meta
-      property="og:title"
-      content="PhyloBioBixBench-Verified-50 | Bio Benchmark"
-    />
+    <meta property="og:title" content="BIOS Benchmark Results" />
     <meta
       property="og:description"
-      content="Interactive dashboard for PhyloBioBixBench-Verified-50 benchmark results — direct, MCQ, and refusal-aware grading across 50 biology questions."
+      content="Interactive dashboard for BixBench-Verified-50 benchmark results — direct, MCQ, and refusal-aware grading across 50 biology questions."
     />
     <meta
       property="og:url"
@@ -30,13 +27,10 @@
 
     <!-- Twitter Card -->
     <meta name="twitter:card" content="summary" />
-    <meta
-      name="twitter:title"
-      content="PhyloBioBixBench-Verified-50 | Bio Benchmark"
-    />
+    <meta name="twitter:title" content="BIOS Benchmark Results" />
     <meta
       name="twitter:description"
-      content="Interactive dashboard for PhyloBioBixBench-Verified-50 benchmark results — direct, MCQ, and refusal-aware grading across 50 biology questions."
+      content="Interactive dashboard for BixBench-Verified-50 benchmark results — direct, MCQ, and refusal-aware grading across 50 biology questions."
     />
 
     <link rel="preconnect" href="https://fonts.googleapis.com" />
@@ -55,17 +49,62 @@
       <!-- 1. Hero -->
       <header class="hero">
         <p class="eyebrow">
-          <a href="https://ai.bio.xyz/" class="eyebrow-link">ai.bio.xyz</a>
+          <a
+            href="https://ai.bio.xyz/"
+            class="eyebrow-link"
+            target="_blank"
+            rel="noopener noreferrer"
+            >ai.bio.xyz</a
+          >
         </p>
-        <h1>PhyloBioBixBench-Verified-50</h1>
+        <h1>BixBench-Verified-50</h1>
         <div class="hero-meta">
           <span id="generated-at">Loading data...</span>
           <span id="totals-badge"></span>
         </div>
         <nav class="hero-links">
-          <a href="https://ai.bio.xyz/" class="hero-link">ai.bio.xyz</a>
-          <a href="https://github.com/bio-xyz/bio-benchmark" class="hero-link"
-            >GitHub</a
+          <a
+            href="https://ai.bio.xyz/"
+            class="hero-link"
+            target="_blank"
+            rel="noopener noreferrer"
+            ><span class="hero-link-icon" aria-hidden="true">
+              <svg viewBox="0 0 24 24" focusable="false">
+                <circle cx="12" cy="12" r="9"></circle>
+                <path d="M3 12h18"></path>
+                <path d="M12 3c2.8 2.4 4.5 5.6 4.5 9s-1.7 6.6-4.5 9"></path>
+                <path d="M12 3c-2.8 2.4-4.5 5.6-4.5 9s1.7 6.6 4.5 9"></path>
+              </svg>
+            </span>
+            <span class="hero-link-label">ai.bio.xyz</span></a
+          >
+          <a
+            href="https://github.com/bio-xyz/bio-benchmark"
+            class="hero-link"
+            target="_blank"
+            rel="noopener noreferrer"
+            ><span class="hero-link-icon" aria-hidden="true">
+              <svg viewBox="0 0 24 24" focusable="false">
+                <path
+                  d="M9 19c-5 1.5-5-2.5-7-3m14 6v-3.87a3.36 3.36 0 0 0-.94-2.61c3.14-.35 6.44-1.54 6.44-7A5.44 5.44 0 0 0 20 4.77 5.07 5.07 0 0 0 19.91 1S18.73.65 16 2.48a13.38 13.38 0 0 0-7 0C6.27.65 5.09 1 5.09 1A5.07 5.07 0 0 0 5 4.77a5.44 5.44 0 0 0-1.5 3.78c0 5.42 3.3 6.61 6.44 7A3.36 3.36 0 0 0 9 18.13V22"
+                ></path>
+              </svg>
+            </span>
+            <span class="hero-link-label">GitHub</span></a
+          >
+          <a
+            href="https://huggingface.co/datasets/phylobio/BixBench-Verified-50"
+            class="hero-link"
+            target="_blank"
+            rel="noopener noreferrer"
+            ><span class="hero-link-icon" aria-hidden="true">
+              <svg viewBox="0 0 24 24" focusable="false">
+                <path d="M4 19h16"></path>
+                <path d="M6 15.5 10.5 11l3.5 3.5L19 8"></path>
+                <path d="m16 8 3-.2-.2 3"></path>
+              </svg>
+            </span>
+            <span class="hero-link-label">Hugging Face</span></a
           >
         </nav>
       </header>
@@ -83,13 +122,19 @@ <h2>Grading Modes</h2>
         aria-label="Headline metrics"
       ></section>
 
-      <!-- 4. Key Takeaways -->
+      <!-- 4. Scientific Reporting -->
+      <section class="insight-panel" id="reporting-section">
+        <h2>Scientific Reporting</h2>
+        <div id="reporting-content" class="reporting-grid"></div>
+      </section>
+
+      <!-- 5. Key Takeaways -->
       <section class="insight-panel" id="takeaways-section">
         <h2>Key Takeaways</h2>
         <ol id="takeaways-list" class="takeaway-list"></ol>
       </section>
 
-      <!-- 5. Source Runs + Downloads -->
+      <!-- 6. Source Runs + Downloads -->
       <section class="panel source-panel">
         <div>
           <h2>Scoped Source Runs</h2>
@@ -111,7 +156,7 @@ <h2>Downloads</h2>
         </div>
       </section>
 
-      <!-- 6. Performance Image -->
+      <!-- 7. Performance Image -->
       <section class="panel">
         <h2>Existing Performance Figure</h2>
         <p class="panel-intro">
@@ -125,7 +170,7 @@ <h2>Existing Performance Figure</h2>
         />
       </section>
 
-      <!-- 7. Charts Grid (6 existing + 1 new task group chart) -->
+      <!-- 8. Charts Grid (6 existing + 1 new task group chart) -->
       <section class="charts">
         <article class="panel chart-panel">
           <h2>Overall Accuracy</h2>
@@ -179,7 +224,7 @@ <h2>Task Group Accuracy (32 groups)</h2>
         </article>
       </section>
 
-      <!-- 8. Strengths & Weaknesses -->
+      <!-- 9. Strengths & Weaknesses -->
       <section class="insight-panel" id="sw-section">
         <h2>Strengths &amp; Weaknesses</h2>
         <div class="sw-grid">
@@ -194,7 +239,7 @@ <h3>Weaknesses</h3>
         </div>
       </section>
 
-      <!-- 9. All 50 Questions Table -->
+      <!-- 10. All 50 Questions Table -->
       <section class="panel">
         <h2>All 50 Questions</h2>
         <div class="table-scroll">
@@ -215,7 +260,7 @@ <h2>All 50 Questions</h2>
         </div>
       </section>
 
-      <!-- 10. All 32 Task Groups Table -->
+      <!-- 11. All 32 Task Groups Table -->
       <section class="panel">
         <h2>All 32 Task Groups</h2>
         <div class="table-scroll">
@@ -235,13 +280,13 @@ <h2>All 32 Task Groups</h2>
         </div>
       </section>
 
-      <!-- 11. MCQ Rescue Detail -->
+      <!-- 12. MCQ Rescue Detail -->
       <section class="panel" id="rescue-detail-section">
         <h2>MCQ Rescue Detail</h2>
         <div id="rescue-detail-content"></div>
       </section>
 
-      <!-- 12. Cross-Run Variability -->
+      <!-- 13. Cross-Run Variability -->
       <section class="panel" id="variability-section">
         <h2>Cross-Run Variability</h2>
         <p class="panel-intro">
@@ -256,12 +301,6 @@ <h2>Cross-Run Variability</h2>
         </div>
       </section>
 
-      <!-- 13. Scientific Reporting -->
-      <section class="insight-panel" id="reporting-section">
-        <h2>Scientific Reporting</h2>
-        <div id="reporting-content" class="reporting-grid"></div>
-      </section>
-
       <!-- 14. Best Repeats Table (all 11) -->
       <section class="panel">
         <h2>Best Repeats</h2>
diff --git a/docs/styles.css b/docs/styles.css
@@ -123,13 +123,14 @@ h1 {
 .hero-links {
   margin-top: 16px;
   display: flex;
+  flex-wrap: wrap;
   gap: 12px;
 }
 
 .hero-link {
   display: inline-flex;
   align-items: center;
-  gap: 6px;
+  gap: 8px;
   padding: 6px 14px;
   border-radius: 8px;
   border: 1px solid var(--border);
@@ -145,6 +146,25 @@ h1 {
   color: var(--accent-green);
 }
 
+.hero-link-icon {
+  display: inline-flex;
+  align-items: center;
+  justify-content: center;
+  width: 14px;
+  height: 14px;
+  flex-shrink: 0;
+}
+
+.hero-link-icon svg {
+  width: 100%;
+  height: 100%;
+  fill: none;
+  stroke: currentColor;
+  stroke-width: 1.8;
+  stroke-linecap: round;
+  stroke-linejoin: round;
+}
+
 code {
   padding: 1px 6px;
   border-radius: 6px;
