Update README.md

README.md

@@ -50,7 +50,7 @@ In our case with NoobAI and Animagine, as the loss landscape is highly non-linea
 To this end, we merge the models according to the weighted average equation used in the Fisher-weighted averaging paper:
 
 $$
-\alpha_i(\lambda) = \frac{\lambda s_{B, i}}{s_{A, i} + s_{B,i}} \\
+\alpha_i(\lambda) = \frac{\lambda s_{B, i}}{(1 - \lambda) s_{A, i} + \lambda s_{B,i}} \\
 \theta_{M, i}(\lambda) = (1 - \alpha_i(\lambda)) \theta_{A, i} + \alpha_i(\lambda) \theta_{B, i}
 $$