Model card updated after epoch 22

README.md

@@ -15,6 +15,6 @@ This repository contains weights for an experimental HRM Causal LM trained on th
 - **Vocab Size**: 32100
 - **Objective:** Causal Language Modeling
 
-### Latest Performance (Epoch 21)
-- **Validation Loss**: `4.6498`
-- **Validation Perplexity**: `104.56`
+### Latest Performance (Epoch 22)
+- **Validation Loss**: `4.6353`
+- **Validation Perplexity**: `103.06`