Update README.md

README.md

@@ -327,11 +327,10 @@ Training Scope	Only LoRA weights updated; main model remains fixed
 
 This approach enables self-corrective, explainable, and meta-aware learning, pushing beyond standard RLHF and toward autonomous reasoning agents.
 
-kl divergence chart under lora gmpo criticism (improvement on the chart)
 <p align="center">
   <img src="https://huggingface.co/liberalusa/liberalmind_bin/resolve/main/kl_critic_plot.png" width="600"/>
 </p>
 
-the block diagram of this technology
 <p align="center">
-  <img src="https://huggingface.co/liberalusa/liberalmind_bin/blob/main/lora_training_diagramab.png" width="400"/></p>
+  <img src="https://huggingface.co/liberalusa/liberalmind_bin/resolve/main/lora_training_diagramab.png" width="400"/>
+</p>