chore: update blog to MiniCPM4.1-8B fine-tune + Sharing is Caring badge earned

BLOG.md

@@ -26,7 +26,7 @@ Everything runs on free infrastructure with zero external API calls. No OpenAI k
 4. **A score ring** tracks your session in real time.
 5. **End the session** and FLUX.2-klein generates a unique image inspired by the topics you just studied — a small visual reward for showing up.
 
-The whole question-answer-feedback loop runs on **MiniCPM4-8B**, an open-weight 8B model, loaded once and shared between question generation and answer evaluation.
+The whole question-answer-feedback loop runs on **MiniCPM4.1-8B**, our QLoRA fine-tune of openbmb's latest 8B model, loaded once and shared between question generation and answer evaluation.
 
 ```
 PDF upload
@@ -153,7 +153,7 @@ Small prompt details made the difference between "tech demo" and "usable study t
 
 | Layer | Choice |
 |---|---|
-| Q&A + evaluation | MiniCPM4-8B (openbmb), bfloat16, transformers 4.57.1 |
+| Q&A + evaluation | MiniCPM4.1-8B (openbmb), QLoRA fine-tune, bfloat16, transformers 4.57.1 |
 | Session images | FLUX.2-klein-4B (Black Forest Labs), diffusers |
 | PDF parsing | PyMuPDF |
 | Backend / hosting | Gradio 6 on Hugging Face Spaces, ZeroGPU |

blog/index.html

@@ -4,7 +4,7 @@
 <meta charset="UTF-8">
 <meta name="viewport" content="width=device-width, initial-scale=1.0">
 <title>How We Fought Gradio and Won. PaperProf Field Notes</title>
-<meta name="description" content="Field notes from the Build Small Hackathon: building PaperProf, an AI study buddy running MiniCPM4-8B and FLUX.2-klein on ZeroGPU with a fully custom UI over a hidden-Gradio bridge.">
+<meta name="description" content="Field notes from the Build Small Hackathon: building PaperProf, an AI study buddy running MiniCPM4.1-8B and FLUX.2-klein on ZeroGPU with a fully custom UI over a hidden-Gradio bridge.">
 <meta property="og:title" content="PaperProf Field Notes. How We Fought Gradio and Won">
 <meta property="og:description" content="68 commits, 10 days, zero external APIs. What we built and what we learned.">
 <meta property="og:type" content="article">
@@ -405,15 +405,15 @@ footer { position: relative; z-index: 1; max-width: 1060px; margin: 0 auto; padd
   <div class="hero-meta">
     <span>READ <b>9 min</b></span>
     <span>TEAM <b>PaperProf · EPITA</b></span>
-    <span>STACK <b>MiniCPM4-8B / FLUX.2 / ZeroGPU</b></span>
+    <span>STACK <b>MiniCPM4.1-8B fine-tune / FLUX.2 / ZeroGPU</b></span>
   </div>
   <span class="scroll-hint">SCROLL</span>
 </header>
 
 <div class="marquee" aria-hidden="true">
   <div class="marquee-track">
-    <span>MiniCPM4-8B</span><span>FLUX.2-klein</span><span>ZeroGPU</span><span>PyMuPDF</span><span>Gradio 6</span><span>Zero external APIs</span><span>68 commits</span><span>10 days</span>
-    <span>MiniCPM4-8B</span><span>FLUX.2-klein</span><span>ZeroGPU</span><span>PyMuPDF</span><span>Gradio 6</span><span>Zero external APIs</span><span>68 commits</span><span>10 days</span>
+    <span>MiniCPM4.1-8B fine-tune</span><span>FLUX.2-klein</span><span>ZeroGPU</span><span>PyMuPDF</span><span>Gradio 6</span><span>Zero external APIs</span><span>68 commits</span><span>10 days</span>
+    <span>MiniCPM4.1-8B fine-tune</span><span>FLUX.2-klein</span><span>ZeroGPU</span><span>PyMuPDF</span><span>Gradio 6</span><span>Zero external APIs</span><span>68 commits</span><span>10 days</span>
   </div>
 </div>
 
@@ -436,13 +436,13 @@ footer { position: relative; z-index: 1; max-width: 1060px; margin: 0 auto; padd
     <div class="badge earned reveal">
       <span class="badge-icon">🔌</span>
       <strong class="badge-name">Off the Grid</strong>
-      <span class="badge-desc">Zero external APIs. MiniCPM4-8B and FLUX.2-klein run entirely on ZeroGPU — no OpenAI key, no rate limits, no data leaving the machine.</span>
+      <span class="badge-desc">Zero external APIs. MiniCPM4.1-8B and FLUX.2-klein run entirely on ZeroGPU — no OpenAI key, no rate limits, no data leaving the machine.</span>
       <span class="badge-chip">✓ Earned</span>
     </div>
     <div class="badge earned reveal">
       <span class="badge-icon">🎯</span>
       <strong class="badge-name">Well-Tuned</strong>
-      <span class="badge-desc">QLoRA fine-tune on SQuAD via Modal. Model published on Hugging Face at <code>build-small-hackathon/MiniCPM4-8B-PaperProf</code>.</span>
+      <span class="badge-desc">QLoRA fine-tune on SQuAD + SciQ via Modal. Model published at <code>build-small-hackathon/MiniCPM4.1-8B-PaperProf</code>.</span>
       <span class="badge-chip">✓ Earned</span>
     </div>
     <div class="badge earned reveal">
@@ -454,7 +454,7 @@ footer { position: relative; z-index: 1; max-width: 1060px; margin: 0 auto; padd
     <div class="badge earned reveal">
       <span class="badge-icon">🦙</span>
       <strong class="badge-name">Llama Champion</strong>
-      <span class="badge-desc">GGUF published at <code>build-small-hackathon/MiniCPM4-8B-PaperProf-GGUF</code>. llama.cpp CPU runtime wired in via <code>PAPERPROF_RUNTIME=llamacpp</code>.</span>
+      <span class="badge-desc">GGUF published at <code>build-small-hackathon/MiniCPM4.1-8B-PaperProf-GGUF</code>. llama.cpp CPU runtime wired in via <code>PAPERPROF_RUNTIME=llamacpp</code>.</span>
       <span class="badge-chip">✓ Earned</span>
     </div>
     <div class="badge earned reveal">
@@ -536,7 +536,7 @@ footer { position: relative; z-index: 1; max-width: 1060px; margin: 0 auto; padd
     <div class="card"><div class="glare"></div><span class="idx">REWARD</span><h4>Session image</h4><p>End the session and FLUX.2-klein generates a unique image from the topics you just studied.</p></div>
   </div>
 
-  <p class="reveal">The whole loop runs on <strong>MiniCPM4-8B</strong>, an open-weight 8B model loaded once and shared between question generation and answer evaluation. PyMuPDF extracts the text, a chunker splits it into thematic sections, and the model picks up from there.</p>
+  <p class="reveal">The whole loop runs on <strong>MiniCPM4.1-8B</strong>, our QLoRA fine-tune of openbmb's latest 8B model, loaded once and shared between question generation and answer evaluation. PyMuPDF extracts the text, a chunker splits it into thematic sections, and the model picks up from there.</p>
 
   <section class="chapter reveal" id="real-story">
     <span class="ghost" aria-hidden="true">03</span>
@@ -553,9 +553,9 @@ footer { position: relative; z-index: 1; max-width: 1060px; margin: 0 auto; padd
 
   <h3 class="reveal" id="lesson-1">Model choice is a compatibility problem, not a benchmark problem</h3>
 
-  <p class="reveal">We started with MiniCPM3-4B, upgraded to MiniCPM4-8B for better reasoning, and immediately hit the classic open-model trap: the model card says one thing, the <code>transformers</code> version on your machine says another.</p>
+  <p class="reveal">We started with MiniCPM3-4B, upgraded to MiniCPM4.1-8B for better reasoning, and immediately hit the classic open-model trap: the model card says one thing, the <code>transformers</code> version on your machine says another.</p>
 
-  <span class="commit reveal">fix: pin transformers==4.57.1 for MiniCPM4-8B compatibility</span>
+  <span class="commit reveal">fix: pin transformers==4.57.1 for MiniCPM4.1-8B compatibility</span>
 
   <p class="reveal">The follow-up lesson came from quantization. Bitsandbytes 4-bit is great on a 16 GB local GPU and completely unnecessary on ZeroGPU hardware, so we made it conditional:</p>
 
@@ -660,7 +660,7 @@ footer { position: relative; z-index: 1; max-width: 1060px; margin: 0 auto; padd
     <table>
       <thead><tr><th scope="col">Layer</th><th scope="col">Choice</th></tr></thead>
       <tbody>
-        <tr><td><strong>Q&amp;A + evaluation</strong></td><td>MiniCPM4-8B (openbmb) · bfloat16 · transformers 4.57.1</td></tr>
+        <tr><td><strong>Q&amp;A + evaluation</strong></td><td>MiniCPM4.1-8B · QLoRA fine-tune (build-small-hackathon/MiniCPM4.1-8B-PaperProf) · bfloat16 · transformers 4.57.1</td></tr>
         <tr><td><strong>Session images</strong></td><td>FLUX.2-klein-4B (Black Forest Labs) · diffusers</td></tr>
         <tr><td><strong>PDF parsing</strong></td><td>PyMuPDF</td></tr>
         <tr><td><strong>Backend / hosting</strong></td><td>Gradio 6 on Hugging Face Spaces · ZeroGPU</td></tr>
@@ -682,7 +682,7 @@ footer { position: relative; z-index: 1; max-width: 1060px; margin: 0 auto; padd
 
 <footer>
   <span>Build Small Hackathon · June 5–15, 2026</span>
-  <span>MiniCPM4-8B / FLUX.2-klein / ZeroGPU / 0 external APIs</span>
+  <span>MiniCPM4.1-8B / FLUX.2-klein / ZeroGPU / 0 external APIs</span>
 </footer>
 
 <script>