Update app.py

app.py

@@ -282,10 +282,9 @@ This tool merges LoRA (Low-Rank Adaptation) adapters with base models using a me
 
 DETAILS_TEXT = """
 ### How It Works
-LoRA enables efficient fine-tuning by adding small adapter weights rather than modifying the entire model. This tool applies additive merging:
+LoRA enables efficient fine-tuning by adding small adapter weights rather than modifying the entire model:
 
-- **Additive LoRA**: `W_new = W + scale × B @ A`
-- Negative scale values reverse the effect
+`W_new = W + scale × (α/r) × B @ A`
 
 ### Scale Factor
 The scale factor (-2 ≤ scale ≤ 2) controls the strength of the LoRA merge: