Update README.md

README.md

@@ -149,7 +149,7 @@ The control vectors in this repository were created experimentally by quadruplin
 
 </details>
 
-Consequently, each model's cross-covariance matrix is now derived from `120,000` hidden state samples. For the largest models (with a hidden dimension of `12,288`), this ensures at least 10 samples per element.
+Consequently, each model's cross-covariance matrix is now derived from `120,000` hidden state samples. For the largest models (with a hidden dimension of `12,288`), this ensures at least ~10 samples per element.
 
 ## Regularisation