docs: publish clean model card (validation + control telemetry)

README.md

@@ -71,10 +71,10 @@ Set your target information ratio \( S^* \), and our PI controller automatically
 <summary><b>Training curves (details)</b></summary>
 
 **DataBPT (bits/token)**  
-![DataBPT curve](assets/figures/data_bpt_curve.png)
+![DataBPT curve](https://raw.githubusercontent.com/Hmbown/shannon-control-unit/main/assets/figures/data_bpt_curve.png)
 
 **ParamBPT (bits/token)**  
-![ParamBPT curve](assets/figures/param_bpt_curve.png)
+![ParamBPT curve](https://raw.githubusercontent.com/Hmbown/shannon-control-unit/main/assets/figures/param_bpt_curve.png)
 
 </details>
 
@@ -129,4 +129,4 @@ Optimal $S^*$ scaling laws are still being discovered. We found 1.0% works for 1
 * **SCU training code:** Apache-2.0 License ([GitHub repository](https://github.com/Hmbown/shannon-control-unit))
 * **IP status:** U.S. patent pending (provisional filed September 2025)
 
-> Repro tips: block size 1024, batch 1, grad-accum 4, gradient checkpointing on, `use_cache=False`.
+> Repro tips: block size 1024, batch 1, grad-accum 4, gradient checkpointing on, `use_cache=False`.