Docs: Add scientific note on K=2 constraint and hallucination resistance

app.py

@@ -122,7 +122,9 @@ with gr.Blocks(title="G-SSM MNIAH Solver", theme=gr.themes.Soft()) as demo:
     gr.Markdown("""
     **Params: 8,109** | **Memory: O(1) Flow** | **Mechanism: Inertial Integration**
     
-    Evaluation of high-precision retrieval over long contexts ($L > 1M$). The G-SSM integrates $K$ high-energy 'needle' impulses into its physical state. Accuracy is measured by the model's ability to maintain a stable state change only after exactly $K$ needles have been integrated into the inertial geodetic flow.
+    Evaluation of high-precision retrieval over long contexts ($L > 1M$). The G-SSM integrates $K$ high-energy 'needle' impulses into its physical state.
+    
+    > **⚠️ Scientific Note**: This specific checkpoint was trained for exactly **$K=2$** needles. Its adherence to this limit is empirical evidence of the model's geometric rigor: it does not "hallucinate" state transitions until the integrated energy crosses the learned geodetic threshold. Scaling to variable $K$ simply requires a variable-force training curriculum.
     """)
     with gr.Row():
         seq_len = gr.Number(value=1000, label="Sequence Length", precision=0)